No BS, just a concise description of exactly what I need to write my own agent.
Alifatisk 5 minutes ago [-]
You might enjoy Z.ais api docs aswell
revolvingthrow 47 minutes ago [-]
> pricing "Pro" $3.48 / 1M output tokens vs $4.40
I’d like somebody to explain to me how the endless comments of "bleeding edge labs are subsidizing the inference at an insane rate" make sense in light of a humongous model like v4 pro being $4 per 1M. I’d bet even the subscriptions are profitable, much less the API prices.
edit: $1.74/M input
$3.48/M output on OpenRouter
schneehertz 23 minutes ago [-]
This price is high even because of the current shortage of inference cards available to DeepSeek; they claimed in their press release that once the Ascend 950 computing cards are launched in the second half of the year, the price of the Pro version will drop significantly
m00x 21 minutes ago [-]
They are profitable to opex costs, but not capex costs with the current depreciation schedules, though those are now edging higher than expected.
mirzap 24 minutes ago [-]
My thoughts exactly. I also believe that subscription services are profitable, and the talk about subsidies is just a way to extract higher profit margins from the API prices businesses pay.
raincole 9 minutes ago [-]
Insert always has been meme.
But seriously, it just stems from the fact some people want AI to go away. If you set your conclusion first, you can very easily derive any premise. AI must go away -> AI must be a bad business -> AI must be losing money.
zarzavat 4 minutes ago [-]
Before the AI bubble that will burst any time now, there was the AI winter that would magically arrive before the models got good enough to rival humans.
masafej536 23 minutes ago [-]
Point taken but there isnt any western providers there yet. Power is cheaper in china.
3uler 11 minutes ago [-]
These models are open and there are tons of western providers offering it at comparable rates.
NitpickLawyer 17 minutes ago [-]
As this is a new arch with tons of optimisations, it'll take some time for inference engines to support it properly, and we'll see more 3rd party providers offer it. Once that settles we'll have a median price for an optimised 1.6T model, and can "guesstimate" from there what the big labs can reasonably serve for the same price. But yeah, it's been said for a while that big labs are ok on API costs. The only unknown is if subscriptions were profitable or not. They've all been reducing the limits lately it seems.
primaprashant 3 minutes ago [-]
While SWE-bench Verified is not a perfect benchmark for coding, AFAIK, this is the first open-weights model that has crossed the threshold of 80% score on this by scoring 80.6%.
Back in Nov 2025, Opus 4.5 (80.9%) was the first proprietary model to do so.
fblp 3 hours ago [-]
There's something heartwarming about the developer docs being released before the flashy press release.
onchainintel 3 hours ago [-]
Insert obligatory "this is the way" Mando scene. Indeed!
necovek 2 hours ago [-]
Where's the training data and training scripts since you are calling this open source?
Edit: it seems "open source" was edited out of the parent comment.
b65e8bee43c2ed0 1 hours ago [-]
doesn't it get tiring after a while? using the same (perceived) gotcha, over and over again, for three years now?
no one is ever going to release their training data because it contains every copyrighted work in existence. everyone, even the hecking-wholesome safety-first Anthropic, is using copyrighted data without permission to train their models. there you go.
necovek 44 minutes ago [-]
There is an easy fix already in widespread use: "open weights".
It is very much a valuable thing already, no need to taint it with wrong promise.
Though I disagree about being used if it was indeed open source: I might not do it inside my home lab today, but at least Qwen and DeepSeek would use and build on what eg. Facebook was doing with Llama, and they might be pushing the open weights model frontier forward faster.
fragmede 57 minutes ago [-]
it's not a gotcha but people using words in ways others don't like.
0-_-0 26 minutes ago [-]
Weights are the source, training data is the compiler.
injidup 15 minutes ago [-]
You got it the wrong way round. It's more akin to.
1. Training data is the source.
2. Training is compilation/compression.
3. Weights are the compiled source akin to optimized assembly.
However it's an imperfect analogy on so many levels. Nitpick away.
bl4ckneon 44 minutes ago [-]
Aww yes, let me push a couple petabytes to my git repo for everyone to download...
necovek 43 minutes ago [-]
An easier thing would be to say "open weights", yes.
yanis_t 2 hours ago [-]
Already on Openrouter. Pro version is $1.74/m/input, $3.48/m/output, while flash $0.14/m/input, 0.28/m/output.
astrod 2 hours ago [-]
Getting 'Api Error' here :(
Every other model is working fine.
poglet 1 hours ago [-]
Try interacting with it through the website, it will give an error and some explanation on the issue. I had to relax my guardrail settings.
Its on OR - but currently not available on their anthropic endpoint. OR if you read this, pls enable it there! I am using kimi-2.6 with Claude Code, works well, but Deepseek V4 gives an error:
`https://openrouter.ai/api/messages with model=deepseek/deepseek-v4-pro, OR returns
an error because their Anthropic-compat translator doesn't cover V4 yet. The Claude CLI dutifully surfaces that error as "model...does not exist"
sidcool 2 hours ago [-]
Truly open source coming from China. This is heartwarming. I know if the potential ulterior motives.
b65e8bee43c2ed0 30 minutes ago [-]
American companies want a scan of your asshole for the privilege of paying to access their models, and unapologetically admit to storing, analyzing, training on, and freely giving your data to any authorities if requested. Chinese ulteriority is hypothetical, American is blatant.
elefanten 18 minutes ago [-]
It’s not remotely hypothetical you’d have to be living under a rock to believe that. And the fusion with a one-party state government that doesn’t tolerate huge swathes of thoughtspace being freely discussed is completely streamlined, not mediated by any guardrails or accountability.
This “no harm to me” meme about a foreign totalitarian government (with plenty of incentive to run influence ops on foreigners) hoovering your data is just so mind-bogglingly naive.
ben_w 1 minutes ago [-]
As a non-American, everything you wrote other than "one party" applies to the current US regime.
Relatively speaking, DeepSeek is less untrustworthy than Grok.
When I try ChatGPT on current events from the White House it interprets them as strange hypotheticals rather than news, which is probably more a problem with DC than with GPT, but whatever.
danny_codes 13 minutes ago [-]
It’s an open model? So you can run it yourself if you want to
theshackleford 5 minutes ago [-]
> This “no harm to me” meme about a foreign totalitarian government (with plenty of incentive to run influence ops on foreigners) hoovering your data is just so mind-bogglingly naive.
This is why I’ve been urging everyone I know to move away from American based services and providers. It’s slow but honest work.
t0lo 13 minutes ago [-]
And you're saying Americans aren't banned from criticising their elites?
tommica 10 minutes ago [-]
Pretty sure you guys have a strong laws about free-speech, and criticizing elites is part of that. Though there are some groups that do not really want the 1st amendment to be a thing.
> Internet comments say that open sourcing is a national strategy, a loss maker subsidized by the government. On the contrary, it is a commercial strategy and the best strategy available in this industry.
This sounds whole lot like potatoh potahto. I think the former argument is very much the correct one: China can undercut everyone and win, even at a loss. Happened with solar panels, steel, evs, sea food - it's a well tested strategy and it works really well despite the many flavors it comes in.
That being said a job well done for the wrong reasons is still a job well done so we should very much welcome these contributions, and maybe it's good to upset western big tech a bit so it's remains competitive.
I_am_tiberius 2 hours ago [-]
Open weight!
alecco 1 hours ago [-]
Please don't slander the most open AI company in the world. Even more open than some non-profit labs from universities. DeepSeek is famous for publishing everything. They might take a bit to publish source code but it's almost always there. And their papers are extremely pro-social to help the broader open AI community. This is why they struggle getting funded because investors hate openness. And in China they struggle against the political and hiring power of the big tech companies.
And DeepSeek often has very cool new approaches to AI copied by the rest. Many others copied their tech. And some of those have 10x or 100x the GPU training budget and that's their moat to stay competitive.
DeepSeek's models are indeed open weight. Why do you feel that pointing this out would be considered slander?
0-_-0 25 minutes ago [-]
Weights are the source, training data is the compiler
crazylogger 14 minutes ago [-]
Training data == source code, training algorithm == compiler, model weights == compiled binary.
mchusma 2 hours ago [-]
For comparison on openrouter DeepSeek v4 Flash is slightly cheaper than Gemma 4 31b, more expensive than Gemma 4 26b, but it does support prompt caching, which means for some applications it will be the cheapest. Excited to see how it compares with Gemma 4.
Model was released and it's amazing. Frontier level (better than Opus 4.6) at a fraction of the cost.
0xbadcafebee 2 hours ago [-]
I don't think we need to compare models to Opus anymore. Opus users don't care about other models, as they're convinced Opus will be better forever. And non-Opus users don't want the expense, lock-in or limits.
As a non-Opus user, I'll continue to use the cheapest fastest models that get my job done, which (for me anyway) is still MiniMax M2.5. I occasionally try a newer, more expensive model, and I get the same results. I have a feeling we might all be getting swindled by the whole AI industry with benchmarks that just make it look like everything's improving.
versteegen 1 hours ago [-]
Which model's best depends on how you use it. There's a huge difference in behaviour between Claude and GPT and other models which makes some poor substitutes for others in certain use cases. I think the GPT models are a bad substitute for Claude ones for tasks such as pair-programming (where you want to see the CoT and have immediate responses) and writing code that you actually want to read and edit yourself, as opposed to just letting GPT run in the background to produce working code that you won't inspect. Yes, GPT 5.4 is cheap and brilliant but very black-box and often very slow IME. GPT-5.4 still seems to behave the same as 5.1, which includes problems like: doesn't show useful thoughts, can think for half an hour, says "Preparing the patch now" then thinks for another 20 min, gives no impression of what it's doing, reads microscopic parts of source files and misses context, will do anything to pass the tests including patching libraries...
ind-igo 1 hours ago [-]
Agree with your assessment, I think after models reached around Opus 4.5 level, its been almost indistinguishable for most tasks. Intelligence has been commoditized, what's important now is the workflows, prompting, and context management. And that is unique to each model.
sandos 25 minutes ago [-]
Is Opus nerfed somehow in Copilot? Ive tried it numerous times, it has never reallt woved me. They seem to have awfully small context windows, but still. Its mostly their reasoning which has been off
Codex is just so much better, or the genera GPT models.
kmarc 1 hours ago [-]
This resonates with me a lot.
I do some stuff with gemini flash and Aider, but mostly because I want to avoid locking myself into a walled garden of models, UIs and company
post-it 1 hours ago [-]
What do you run these on? I've gotten comfortable with Claude but if folks are getting Opus performance for cheaper I'll switch.
oceanplexian 18 minutes ago [-]
You can just use Claude Code with a few env vars, most of these providers offer an Anthropic compatible API
slopinthebag 1 hours ago [-]
Try Charm Crush first, it's a native binary. If it's unbearable, try opencode, just with the knowledge your system will probably be pwned soon since it's JS + NPM + vibe coding + some of the most insufferable devs in the industry behind that product.
If you're feeling frisky, Zed has a decent agent harness and a very good editor.
sandGorgon 37 minutes ago [-]
actually this is not the reason - the harness is significantly better.
There is no comparable harness to Claude Code with skills, etc.
Opencode was getting there, but it seems the founders lost interest. Pi could be it, but its very focused on OpenClaw. Even Codex cli doesnt have all of it.
which harness works well with Deepseek v4 ?
darkwater 20 minutes ago [-]
What's the issue with OC? I tried it a bit over 2 months ago, when I was still on Claude API, and it actually liked more that CC (i.e. the right sidebar with the plan and a tendency at asking less "security" questions that CC). Why is it so bad nowadays?
avereveard 26 minutes ago [-]
eh idk. until yesterday opus was the one that got spatial reasoning right (had to do some head pose stuff, neither glm 5.1 nor codex 5.3 could "get" it) and codex 5.3 was my champion at making UX work.
So while I agree mixed model is the way to go, opus is still my workhorse.
szundi 1 hours ago [-]
[dead]
onchainintel 3 hours ago [-]
How does it compare to Opus 4.7? I've been immersed in 4.7 all week participating in the Anthropic Opus 4.7 hackathon and it's pretty impressive even if it's ravenous from a token perspective compared to 4.6
greenknight 2 hours ago [-]
The thing is, it doesnt need to beat 4.7. it just needs to do somewhat well against it.
This is free... as in you can download it, run it on your systems and finetune it to be the way you want it to be.
libraryofbabel 38 minutes ago [-]
> you can download it, run it on your systems
In theory, sure, but as other have pointed out you need to spend half a million on GPUs just to get enough VRAM to fit a single instance of the model. And you’d better make sure your use case makes full 24/7 use of all that rapidly-depreciating hardware you just spent all your money on, otherwise your actual cost per token will be much higher than you think.
In practice you will get better value from just buying tokens from a third party whose business is hosting open weight models as efficiently as possible and who make full use of their hardware. Even with the small margin they charge on top you will still come out ahead.
oceanplexian 12 minutes ago [-]
There are a lot of companies who would gladly drop half a million on a GPU to have private inference that Anthropic or OpenAI can’t use to steal their data.
And that GPU wouldn’t run one instance, the models are highly parallelizable. It would likely support 10-15 users at once, if a company oversubscribed 10:1 that GPU supports ~100 seats. Amortized over a couple years the costs are competitive.
hsbauauvhabzb 21 minutes ago [-]
Sure, but that’s an incredibly short term viewpoint.
p1esk 2 hours ago [-]
Do you think a lot of people have “systems” to run a 1.6T model?
CJefferson 1 hours ago [-]
To me, the important thing isn't that I can run it, it's that I can pay someone else to run it. I'm finding Opus 4.7 seems to be weirdly broken compared to 4.6, it just doesn't understand my code, breaks it whenever I ask it to do anything.
Now, at the moment, i can still use 4.6 but eventually Anthropic are going to remove it, and when it's gone it will be gone forever. I'm planning on trying Deepseek v4, because even if it's not quite as good, I know that it will be available forever, I'll always be able to find someone to run it.
applfanboysbgon 2 hours ago [-]
No, but businesses do. Being able to run quality LLMs without your business, or business's private information, being held at the mercy of another corp has a lot of value.
forrestthewoods 2 hours ago [-]
What type of system is needed to self host this? How much would it cost?
disiplus 1 hours ago [-]
Depends how many users you have and what is "production grade" for you but like 500k gets you a 8x B200 machine.
CamperBob2 6 minutes ago [-]
$20K worth of RTX 6000 Blackwell cards should let you run the Flash version of the model.
p1esk 1 hours ago [-]
Depends on fast you want it to be. I’m guessing a couple of $10k mac studio boxes could run it, but probably not fast enough to enjoy using it.
fragmede 1 hours ago [-]
One GB200 NVL72 from Nvidia would do it. $2-3 million, or so. If you're a corporation, say Walmart or PayPal, that's not out of the question.
If you want to go budget corporate, 7 x H200 is just barely going to run it, but all in, $300k ought to do it.
gloflo 1 hours ago [-]
How many users can you serve with that?
fragmede 23 minutes ago [-]
For the H200, between 150-700. The GB200 gets you something like 2-10k users.
choldstare 2 hours ago [-]
Not really - on prem llm hosting is extremely labor and capital intensive
applfanboysbgon 2 hours ago [-]
But can be, and is, done. I work for a bootstrapped startup that hosts a DeepSeek v3 retrain on our own GPUs. We are highly profitable. We're certainly not the only ones in the space, as I'm personally aware of several other startups hosting their own GLM or DeepSeek models.
2 hours ago [-]
onchainintel 2 hours ago [-]
Completely agree, not suggesting it needs ot just genuinely curious. Love that it can be run locally though. Open source LLMs punching back pretty hard against proprietary ones in the cloud lately in terms of performance.
- To run with "heavy quantization" (16 bits -> 8): "8xH100", giving us $200K upfront and $4/h.
- To run truly "locally"--i.e. in a house instead of a data center--you'd need four 4090s, one of the most powerful consumer GPUs available. Even that would clock in around $15k for the cards alone and ~$0.22/h for the electricity (in the US).
Truly an insane industry. This is a good reminder of why datacenter capex from since 2023 has eclipsed the Manhattan Project, the Apollo program, and the US interstate system combined...
oceanplexian 6 minutes ago [-]
All these number are peanuts to a mid sized company. A place I worked at used to spend a couple million just for a support contract on a Netapp.
10 years from now that hardware will be on eBay for any geek with a couple thousand dollars and enough power to run it.
zargon 1 hours ago [-]
That article is a total hallucination.
"671B total / 37B active"
"Full precision (BF16)"
And they claim they ran this non-existent model on vLLM and SGLang over a month and a half ago.
It's clickbait keyword slop filled in with V3 specs. Most of the web is slop like this now. Sigh.
slashdave 2 hours ago [-]
"if you have to ask..."
johnmaguire 2 hours ago [-]
... if you have 800 GB of VRAM free.
inventor7777 2 hours ago [-]
I remember reading about some new frameworks have been coming out to allow Macs to stream weights of huge models live from fast SSDs and produce quality output, albeit slowly. Apart from that...good luck finding that much available VRAM haha
rvz 2 hours ago [-]
It is more than good enough and has effectively caught up with Opus 4.6 and GPT 5.4 according to the benchmarks.
It's about 2 months behind GPT 5.5 and Opus 4.7.
As long as it is cheap to run for the hosting providers and it is frontier level, it is a very competitive model and impressive against the others. I give it 2 years maximum for consumer hardware to run models that are 500B - 800B quantized on their machines.
It should be obvious now why Anthropic really doesn't want you to run local models on your machine.
deaux 2 hours ago [-]
Vibes > Benchmarks. And it's all so task-specific. Gemini 3 has scored very well in benchmarks for very long but is poor at agentic usecases. A lot of people prefering Opus 4.6 to 4.7 for coding despite benchmarks, much more than I've seen before (4.5->4.6, 4->4.5).
Doesn't mean Deepseek v4 isn't great, just benchmarks alone aren't enough to tell.
snovv_crash 2 hours ago [-]
With the ability of the Qwen3.6 27B, I think in 2 years consumers will be running models of this capability on current hardware.
colordrops 2 hours ago [-]
What's going to change in 2 years that would allow users to run 500B-800B parameter models on consumer hardware?
DiscourseFan 2 hours ago [-]
I think its just an estimate
indigodaddy 60 minutes ago [-]
But the question remains
doctoboggan 3 hours ago [-]
Is it honestly better than Opus 4.6 or just benchmaxxed? Have you done any coding with an agent harness using it?
If its coding abilities are better than Claude Code with Opus 4.6 then I will definitely be switching to this model.
bokkies 44 minutes ago [-]
Apparently glm5.1 and qwen coder latest is as good as opus 4.6 on benchmarks. So I tried both seriously for a week (glm Pro using CC) and qwen using qwen companion. Thought I could save $80 a month. Unfortunately after 2 days I had switched back to Max. The speed (slower on both although qwen is much faster) and errors (stupid layout mistakes, inserting 2 footers then refusing to remove one, not seeing obvious problems in screenshots & major f-ups of functionality), not being able to view URLs properly, etc. I'll give deepseek a go but I suspect it will be similar. The model is only half the story. Also been testing gpt5.4 with codex and it is very almost as good as CC... better on long running tasks running in background. Not keen on ChatGPT codex 'personality' so will stick to CC for the most part.
madagang 2 hours ago [-]
Their Chinese announcement says that, based on internal employee testing, it is not as good as Opus 4.6 Thinking, but is slightly better than Opus 4.6 without Thinking enabled.
I appreciate this, makes me trust it more than benchmarks.
deaux 2 hours ago [-]
That's super interesting, isn't Deepseek in China banned from using Anthropic models? Yet here they're comparing it in terms of internal employee testing.
renticulous 25 minutes ago [-]
They use VPN to access. Even Google Deepmind uses Anthropic. There was a fight within Google as to why only DeepMind is allowed to Claude while rest of the Google can't.
2 hours ago [-]
NitpickLawyer 2 hours ago [-]
> (better than Opus 4.6)
There we go again :) It seems we have a release each day claiming that. What's weird is that even deepseek doesn't claim it's better than opus w/ thinking. No idea why you'd say that but anyway.
Dsv3 was a good model. Not benchmaxxed at all, it was pretty stable where it was. Did well on tasks that were ood for benchmarks, even if it was behind SotA.
This seems to be similar. Behind SotA, but not by much, and at a much lower price. The big one is being served (by ds themselves now, more providers will come and we'll see the median price) at 1.74$ in / 3.48$ out / 0.14$ cache. Really cheap for what it offers.
The small one is at 0.14$ in / 0.28$ out / 0.028$ cache, which is pretty much "too cheap to matter". This will be what people can run realistically "at home", and should be a contender for things like haiku/gemini-flash, if it can deliver at those levels.
slopinthebag 1 hours ago [-]
Anthropic fans would claim God itself is behind Opus by 3-6 months and then willingly be abused by Boris and one of his gaslighting tweets.
> According to evaluation feedback, its user experience is better than Sonnet 4.5, and its delivery quality is close to Opus 4.6's non-thinking mode, but there is still a certain gap compared to Opus 4.6's thinking mode.
This is the model creators saying it, not me.
bbor 2 hours ago [-]
For the curious, I did some napkin math on their posted benchmarks and it racks up 20.1 percentage point difference across the 20 metrics where both were scored, for an average improvement of about 2% (non-pp). I really can't decide if that's mind blowing or boring?
Claude4.6 was almost 10pp better at at answering questions from long contexts ("corpuses" in CorpusQA and "multiround conversations" in MRCR), while DSv4 was a staggering 14pp better at one math challenge (IMOAnswerBench) and 12pp better at basic Q&A (SimpleQA-Verified).
Quasimarion 2 hours ago [-]
FWIW it's also like 10x cheaper.
51 minutes ago [-]
sergiotapia 3 hours ago [-]
The dragon awakes yet again!
kindkang2024 2 hours ago [-]
There appears a flight of dragons without heads. Good fortune.
That's literally what the I Ching calls "good fortune."
Competition, when no single dragon monopolizes the sky, brings fortune for all.
rapind 3 hours ago [-]
Pop?
gbnwl 3 hours ago [-]
I’m deeply interested and invested in the field but I could really use a support group for people burnt out from trying to keep up with everything. I feel like we’ve already long since passed the point where we need AI to help us keep up with advancements in AI.
satvikpendem 1 hours ago [-]
Don't keep up. Much like with news, you'll know when you need to know, because someone else will tell you first.
wordpad 3 hours ago [-]
The players barely ever change. People don't have problems following sports, you shouldn't struggle so much with this once you accept top spot changes.
gbnwl 1 hours ago [-]
I didn't express this well but my interest isn't "who is in the top spot", and is more _why and _how various labs get the results they do. This is also magnified by the fact that I'm not only interested in hosted providers of inference but local models as well. What's your take on the best model to run for coding on 24GB of VRAM locally after the last few weeks of releases? Which harness do you prefer? What quants do you think are best? To use your sports metaphor it's more than following the national leagues but also following college and even high school leagues as well. And the real interest isn't even who's doing well but WHY, at each level.
renticulous 21 minutes ago [-]
Follow the AI newsletters. They bundle the news along with their Op-Ed and summarize it better.
ehnto 2 hours ago [-]
It is funny seeing people ping pong between Anthropic and ChatGPT, with similar rhetoric in both directions.
At this point I would just pick the one who's "ethics" and user experience you prefer. The difference in performance between these releases has had no impact on the meaningful work one can do with them, unless perhaps they are on the fringes in some domain.
Personally I am trying out the open models cloud hosted, since I am not interested in being rug pulled by the big two providers. They have come a long way, and for all the work I actually trust to an LLM they seem to be sufficient.
DiscourseFan 2 hours ago [-]
I find ChatGPT annoying mostly
awakeasleep 2 hours ago [-]
Open settings > personalization. Set it to efficient base style. Turn off enthusiasm and warmth. You’re welcome
vrganj 55 minutes ago [-]
It honestly has all kinda felt like more of the same ever since maybe GPT4?
New model comes out, has some nice benchmarks, but the subjective experience of actually using it stays the same. Nothing's really blown my mind since.
Feels like the field has stagnated to a point where only the enthusiasts care.
trueno 1 hours ago [-]
holy shit im right there with you
zkmon 32 minutes ago [-]
They released 1.6 T pro base model on huggingface. First time I'm seeing a "T" model here.
bandrami 35 minutes ago [-]
I don't mind that High Flyer completely ripped off Anthropic to do this so much as I mind that they very obviously waited long enough for the GAB to add several dozen xz-level easter eggs to it.
apexalpha 12 minutes ago [-]
This FLash model might be affordable for OpenClaw. I run it on my mac 48gb ram now but it's slowish.
CJefferson 1 hours ago [-]
What's the current best framework to have a 'claude code' like experience with Deepseek (or in general, an open-source model), if I wanted to play?
Just tested it via openrounter in the Pi Coding agent and it regularly fails to use the read and write tool correctly, very disappointing. Anyone know a fix besides prompting "always use the provided tools instead of writing your own call"
abstracthinking 24 minutes ago [-]
They have just released it, give it some time, they probably haven't pretested it with Pi
Imanari 9 minutes ago [-]
How can they fix it after the release? They would have to retrain/finetune it further, no?
zargon 3 minutes ago [-]
It's only in preview right now. And anyway, yes, models regularly get updated training.
But in this case, it's more likely just to be a prompting or chat template issue.
zargon 2 hours ago [-]
The Flash version is 284B A13B in mixed FP8 / FP4 and the full native precision weights total approximately 154 GB. KV cache is said to take 10% as much space as V3. This looks very accessible for people running "large" local models. It's a nice follow up to the Gemma 4 and Qwen3.5 small local models.
sbinnee 2 hours ago [-]
Price is appealing to me. I have been using gemini 3 flash mainly for chat. I may give it a try.
input: $0.14/$0.28 (whereas gemini $0.5/$3)
Does anyone know why output prices have such a big gap?
tokenmaxxinej 5 minutes ago [-]
input tokens are processed at 10-50 times the speed of output tokens since you can process then in batches and not one at a time like output tokens
girvo 33 minutes ago [-]
Output is what the compute is used for above all else; costs more hardware time basically than prompt processing (input) which is a lot faster
simonw 2 hours ago [-]
I like the pelican I got out of deepseek-v4-flash more than the one I got from deepseek-v4-pro.
This is just a random thought, but have you tried doing an 'agentic' pelican?
As in have the model consider its generated SVG, and gradually refine it, using its knowledge of the relative positions and proportions of the shapes generated, and have it spin for a while, and hopefully the end result will be better than just oneshotting it.
Or maybe going even one step further - most modern models have tool use and image recognition capabilities - what if you have it generate an SVG (or parts/layers of it, as per the model's discretion) and feed it back to itself via image recognition, and then improve on the result.
I think it'd be interesting to see, as for a lot of models, their oneshot capability in coding is not necessarily corellated with their in-harness ability, the latter which really matters.
JSR_FDED 2 hours ago [-]
No way. The Pro pelican is fatter, has a customized front fork, and the sun is shining! He’s definitely living the best life.
chronogram 36 minutes ago [-]
The pro pelican is a work of art! It goes dimensions that no other LLM has gone before.
w4yai 2 hours ago [-]
yeah. look at these 4 feathers (?) on his bum too.
oliver236 1 hours ago [-]
a lot of dumplings
nickvec 2 hours ago [-]
The Flash one is pretty impressive. Might be my favorite so far in the pelican-riding-a-bicycle series
murkt 58 minutes ago [-]
DeepSeek pelicans are the angriest pelicans I’ve seen so far.
kristopolous 56 minutes ago [-]
they're just late for work.
mikae1 1 hours ago [-]
Being a bicycle geometry nerd I always look at the bicycle first.
Let me tell you how much the Pro one sucks... It looks like failed Pedersen[1]. The rear wheel intersects with the bottom bracket, so it wouldn't even roll. Or rather, this bike couldn't exist.
The flash one looks surprisingly correct with some wild fork offset and the slackest of seat tubes. It's got some lowrider[2] aspirations with the small wheels, but with longer, Rivendellish[3], chainstays. The seat post has different angle than the seat tube, so good luck lowering that.
This is an excellent comment. Thanks for this - I've only ever thought about whether the frame is the right shape, I never thought about how different illustrations might map to different bicycle categories.
mikae1 54 minutes ago [-]
Some other reactions:
I wonder which model will try some more common spoke lacing patterns. Right now there seems to be a preference for radial lacing, which is not super common (but simple to draw). The Flash and Pro one uses 16 spoke rims, which actually exist[1] but are not super common.
The Pro model fails badly at the spokes. Heck, the spokes sit on the outside of the drive side of the rim and tire. Have a nice ride riding on the spokes (instead of the tire) welded to the side of your rim.
Both bikes have the drive side on the left, which is very very uncommon. That can't exist in the training data.
The Pedersen looks like someone failed the "draw a bicycle" test and decided to adjust the universe.
nsoonhui 40 minutes ago [-]
To me this is the perfect proof that
1) LLM is not AGI. Because surely if AGI it would imply that pro would do better than flash?
2) and because of the above, Pelican example is most likely already being benchmaxxed.
chvid 38 minutes ago [-]
Is it then Deepseek hosted by Deepseek?
How much does the drawing change if you ask it again?
catelm 56 minutes ago [-]
I think the pelican on a bike is known widely enough that of seizes to be useful as a benchmark. There is even a pelican briefly appearing in the promo video of GPT-5, if I'm not mistaken https://openai.com/gpt-5/. So the companies are apparently aware of it.
brutal_chaos_ 53 minutes ago [-]
What was your prompt for the image? Apologies if this should be obvious.
shawn_w 51 minutes ago [-]
>Generate an SVG of a pelican riding a bicycle
at the top of the linked pages.
theanonymousone 59 minutes ago [-]
Where is the GPT 5.5 Pelican?
culopatin 36 minutes ago [-]
In the 5.5 topic
ycui1986 2 hours ago [-]
I really like the pro version. The pelican is so cute.
EnPissant 42 minutes ago [-]
This should not be the top comment on every model release post. It's getting tiring.
blitzar 35 minutes ago [-]
This should be the bottom comment on the pelican comment on every model release post.
EnPissant 24 minutes ago [-]
Clearly the top comment should be "Imagine a beowulf cluster of Deepseek v4!"
ButlerianJihad 21 minutes ago [-]
My mother was murdered by Beowulf, you insensitive Claude!
EnPissant 18 minutes ago [-]
This was perfect.
lobochrome 59 minutes ago [-]
Why they so angry?
whateveracct 1 hours ago [-]
[flagged]
fastball 1 hours ago [-]
It's just Simon Willison (the person you are replying to) who always makes a pelican, as his personal flippant benchmark. It's not that deep.
dewey 1 hours ago [-]
No benchmark will be perfect, especially if it's public but it's a fun experiment to visually see how these models get better and better.
post-it 1 hours ago [-]
Why is it so wrong?
simonw 1 hours ago [-]
Thanks for the "scientific air" remark, that gave me a genuine LOL.
rohanm93 52 minutes ago [-]
This is shockingly cheap for a near frontier model. This is insane.
For context, for an agent we're working on, we're using 5-mini, which is $2/1m tokens. This is $0.30/1m tokens. And it's Opus 4.6 level - this can't be real.
I am uncomfortable about sending user data which may contain PII to their servers in China so I won't be using this as appealing as it sounds. I need this to come to a US-hosted environment at an equivalent price.
Hosting this on my own + renting GPUs is much more expensive than DeepSeek's quoted price, so not an option.
fractalf 42 minutes ago [-]
Right now Im much more worried about sending data to the US and A.. At least theres a less chanse it will be missused against -me-
xnx 40 minutes ago [-]
Such different time now than early 2025 when people thought Deepaeek was going to kill the market for Nvidia.
jessepcc 2 hours ago [-]
At this point 'frontier model release' is a monthly cadence, Kimi 2.6 Claude 4.6 GPT 5.5, the interesting question is which evals will still be meaningful in 6 months.
storus 1 hours ago [-]
Oh well, I should have bought 2x 512GB RAM MacStudios, not just one :(
Aliabid94 3 hours ago [-]
MMLU-Pro:
Gemini-3.1-Pro at 91.0
Opus-4.6 at 89.1
GPT-5.4, Kimi2.6, and DS-V4-Pro tied at 87.5
Pretty impressive
ant6n 2 hours ago [-]
Funny how Gemini is theoretically the best -- but in practice all the bugs in the interface mean I don't want to use it anymore. The worst is it forgets context (and lies about it), but it's very unreliable at reading pdfs (and lies about it). There's also no branch, so once the context is lost/polluted, you have to start projects over and build up the context from scratch again.
clark1013 1 hours ago [-]
Looking forward to DeepSeek Coding Plan
m_abdelfattah 27 minutes ago [-]
I came here to say the same :) !
jdeng 3 hours ago [-]
Excited that the long awaited v4 is finally out. But feel sad that it's not multimodal native.
WhereIsTheTruth 8 minutes ago [-]
Interesting note:
"Due to constraints in high-end compute capacity, the current service capacity for Pro is very limited. After the 950 supernodes are launched at scale in the second half of this year, the price of Pro is expected to be reduced significantly."
So it's going to be even cheaper
tcbrah 26 minutes ago [-]
giving meta a run for its money, esp when it was supposed to be the poster child for OSS models. deepseek is really overshadowing them rn
dang, probably the two should be merged and that be the link
culi 2 hours ago [-]
there's no pinging. Someone's gotta email dang
tariky 1 hours ago [-]
Anyone tried with make web UI with it? How good is it? For me opus is only worth because of it.
sibellavia 57 minutes ago [-]
A few hours after GPT5.5 is wild. Can’t wait to try it.
KaoruAoiShiho 3 hours ago [-]
SOTA MRCR (or would've been a few hours earlier... beaten by 5.5), I've long thought of this as the most important non-agentic benchmark, so this is especially impressive. Beats Opus 4.7 here
luew 54 minutes ago [-]
We will be hosting it soon at getlilac.com!
reenorap 2 hours ago [-]
Which version fits in a Mac Studio M3 Ultra 512 GB?
How can you reasonably try to get near frontier (even at all tps) on hardware you own? Maybe under 5k in cost?
revolvingthrow 1 hours ago [-]
For flash? 4 bit quant, 2x 96GB gpu (fast and expensive) or 1x 96GB gpu + 128GB ram (still expensive but probably usable, if you’re patient).
A mac with 256 GB memory would run it but be very slow, and so would be a 256GB ram + cheapo GPU desktop, unless you leave it running overnight.
The big model? Forget it, not this decade. You can theoretically load from SSD but waiting for the reply will be a religious experience.
Realistically the biggest models you can run on local-as-in-worth-buying-as-a-person hardware are between 120B and 200B, depending on how far you’re willing to go on quantization. Even this is fairly expensive, and that’s before RAM went to the moon.
zargon 58 minutes ago [-]
Flash is less than 160 GB. No need to quantize to fit in 2x 96 GB. Not sure how much context fits in 30 GB, but it should be a good amount.
redrove 41 minutes ago [-]
It seems to be 160GB at mixed FP4+FP8 precision, FYI. Full FP8 is 250GB+. (B)F16 at around double I would assume.
zargon 38 minutes ago [-]
There is no BF16. The instruct model at full precision is 160 GB (mixed FP4 and FP8). The base model at full precision is 284 GB (FP8). Almost everyone is going to use instruct. But I do love to see base models released.
zozbot234 34 minutes ago [-]
Run on an old HEDT platform with a lot of parallel attached storage (probably PCIe 4) and fetch weights from SSD. You'd ultimately be limited by the latency of these per-layer fetches, since MoE weights are small. You could reduce the latencies further by buying cheap Optane memory on the second-hand market.
awakeasleep 2 hours ago [-]
The same way you fit a bucket wheel excavator in your garage
floam 1 hours ago [-]
Very carefully
datadrivenangel 1 hours ago [-]
A loaded macbook pro can get you to the frontier from 24 months ago at ~10-40tok/s, which is plenty fast enough for regular chatting.
542458 1 hours ago [-]
The low end could be something like an eBay-sourced server with a truckload of DDR3 ram doing all-cpu inference - secondhand server models with a terabyte of ram can be had for about 1.5K. The TPS will be absolute garbage and it will sound like a jet engine, but it will nominally run.
The flash version here is 284B A13B, so it might perform OK with a fairly small amount of VRAM for the active params and all regular ram for the other params, but I’d have to see benchmarks. If it turns out that works alright, an eBay server plus a 3090 might be the bang-for-buck champ for about $2.5K (assuming you’re starting from zero).
jdoe1337halo 2 hours ago [-]
More like 500k
mariopt 2 hours ago [-]
Does deepseek has any coding plan?
jeffzys8 1 hours ago [-]
no
swrrt 2 hours ago [-]
Any visualised benchmark/scoreboard for comparison between latest models? DeepSeek v4 and GPT-5.5 seems to be ground breaking.
namegulf 2 hours ago [-]
Is there a Quantized version of this?
sergiotapia 49 minutes ago [-]
Using it with opencode sometimes it generates commands like:
bash({"command":"gh pr create --title "Improve Calendar module docs and clean up idiomatic Elixir" --body "$(cat <<'EOF'
Problem
The Calendar modu...
like generating output, but not actually running the bash command so not creating the PR ultimately. I wonder if it's a model thing, or an opencode thing.
rvz 2 hours ago [-]
The paper is here: [0]
Was expecting that the release would be this month [1], since everyone forgot about it and not reading the papers they were releasing and 7 days later here we have it.
One of the key points of this model to look at is the optimization that DeepSeek made with the residual design of the neural network architecture of the LLM, which is manifold-constrained hyper-connections (mHC) which is from this paper [2], which makes this possible to efficiently train it, especially with its hybrid attention mechanism designed for this.
There was not that much discussion around it some months ago here [3] about it but again this is a recommended read of the paper.
I wouldn't trust the benchmarks directly, but would wait for others to try it for themselves to see if it matches the performance of frontier models.
Either way, this is why Anthropic wants to ban open weight models and I cannot wait for the quantized versions to release momentarily.
> this is why Anthropic wants to ban open weight models
Do you have a source?
louiereederson 1 hours ago [-]
More like he wants to ban accelerator chip sales to China, which may be about “national security” or self preservation against a different model for AI development which also happens to be an existential threat to Anthropic. Maybe those alternatives are actually one and the same to him.
2 hours ago [-]
punkpeye 41 minutes ago [-]
Incredible model quality to price ratio
ls612 2 hours ago [-]
How long does it usually take for folks to make smaller distills of these models? I really want to see how this will do when brought down to a size that will run on a Macbook.
simonw 2 hours ago [-]
Unsloth often turn them around within a few hours, they might have gone to bed already though!
Weren't there some frameworks recently released to allow Macs to stream weights from fast SSDs and thus fit way more parameters than what would normally fit in RAM?
I have never tried one yet but I am considering trying that for a medium sized model.
simonw 2 hours ago [-]
I've been calling that the "streaming experts" trick, the key idea is to take advantage of Mixture of Expert models where only a subset of the weights are used for each round of calculations, then load those weights from SSD into RAM for each round.
As I understand it if DeepSeek v4 Pro is a 1.6T, 49B active that means you'd need just 49B in memory, so ~100GB at 16 bit or ~50GB at 8bit quantized.
v4 Flash is 284B, 13B active so might even fit in <32GB.
zozbot234 41 minutes ago [-]
The "active" count is not very meaningful except as a broad measure of sparsity, since the experts in MoE models are chosen per layer. Once you're streaming experts from disk, there's nothing that inherently requires having 49B parameters in memory at once. Of course, the less caching memory does, the higher the performance overhead of fetching from disk.
zargon 1 hours ago [-]
> ~100GB at 16 bit or ~50GB at 8bit quantized.
V4 is natively mixed FP4 and FP8, so significantly less than that. 50 GB max unquantized.
inventor7777 2 hours ago [-]
Ahh, that actually makes more sense now. (As you can tell, I just skimmed through the READMEs and starred "for later".)
My Mac can fit almost 70B (Q3_K_M) in memory at once, so I really need to try this out soon at maybe Q5-ish.
EnPissant 35 minutes ago [-]
Streaming weights from RAM to GPU for prefill makes sense due to batching and pcie5 x16 is fast enough to make it worthwhile.
Streaming weights from RAM to GPU for decode makes no sense at all because batching requires multiple parallel streams.
Streaming weights from SSD _never_ makes sense because the delta between SSD and RAM is too large. There is no situation where you would not be able to fit a model in RAM and also have useful speeds from SSD.
These are more like experiments than a polished release as of yet. And the reduction in throughput is high compared to having the weights in RAM at all times, since you're bottlenecked by the SSD which even at its fastest is much slower than RAM.
the_sleaze_ 2 hours ago [-]
Do you have the links for those? Very interested
inventor7777 2 hours ago [-]
Sure!
Note: these were just two that I starred when I saw them posted here. I have not looked seriously at it at the moment,
I suspect you may have replied to a bot. Dead internet theory
slopinthebag 1 hours ago [-]
OMG
OMG ITS HAPPENING
shafiemoji 3 hours ago [-]
I hope the update is an improvement. Losing 3.2 would be a real loss, it's excellent.
raincole 2 hours ago [-]
History doesn't always repeat itself.
But if it does, then in the following week we'll see DeepSeek4 floods every AI-related online space. Thousands of posts swearing how it's better than the latest models OpenAI/Anthropic/Google have but only costs pennies.
Then a few weeks later it'll be forgotten by most.
sbysb 2 hours ago [-]
It's difficult because even if the underlying model is very good, not having a pre-built harness like Claude Code makes it very un-sticky for most devs. Even at equal quality, the friction (or at least perceived friction) is higher than the mainstream models.
raincole 2 hours ago [-]
OpenCode? Pi?
If one finds it difficult to set up OpenCode to use whatever providers they want, I won't call them 'dev'.
The only real friction (if the model is actually as good as SOTA) is to convince your employer to pay for it. But again if it really provides the same value at a fraction of the cost, it'll eventually cease to be an issue.
throwa356262 40 minutes ago [-]
"If one finds it difficult to set up OpenCode to use whatever providers they want, I won't call them 'dev'."
I feel the same way. But look at the llama vs llama.cpp post from HN few days back and you will see most of the enthusiasts in this space are very non technical people.
zargon 6 minutes ago [-]
I think you mean ollama vs llama.cpp.
cmrdporcupine 2 hours ago [-]
They have instructions right on their page on how to use claude code with it.
slopinthebag 1 hours ago [-]
[flagged]
Rendered at 06:29:16 GMT+0000 (Coordinated Universal Time) with Vercel.
https://api-docs.deepseek.com/guides/thinking_mode
No BS, just a concise description of exactly what I need to write my own agent.
I’d like somebody to explain to me how the endless comments of "bleeding edge labs are subsidizing the inference at an insane rate" make sense in light of a humongous model like v4 pro being $4 per 1M. I’d bet even the subscriptions are profitable, much less the API prices.
edit: $1.74/M input $3.48/M output on OpenRouter
But seriously, it just stems from the fact some people want AI to go away. If you set your conclusion first, you can very easily derive any premise. AI must go away -> AI must be a bad business -> AI must be losing money.
Back in Nov 2025, Opus 4.5 (80.9%) was the first proprietary model to do so.
Edit: it seems "open source" was edited out of the parent comment.
no one is ever going to release their training data because it contains every copyrighted work in existence. everyone, even the hecking-wholesome safety-first Anthropic, is using copyrighted data without permission to train their models. there you go.
It is very much a valuable thing already, no need to taint it with wrong promise.
Though I disagree about being used if it was indeed open source: I might not do it inside my home lab today, but at least Qwen and DeepSeek would use and build on what eg. Facebook was doing with Llama, and they might be pushing the open weights model frontier forward faster.
1. Training data is the source. 2. Training is compilation/compression. 3. Weights are the compiled source akin to optimized assembly.
However it's an imperfect analogy on so many levels. Nitpick away.
https://openrouter.ai/deepseek/deepseek-v4-flash
`https://openrouter.ai/api/messages with model=deepseek/deepseek-v4-pro, OR returns an error because their Anthropic-compat translator doesn't cover V4 yet. The Claude CLI dutifully surfaces that error as "model...does not exist"
This “no harm to me” meme about a foreign totalitarian government (with plenty of incentive to run influence ops on foreigners) hoovering your data is just so mind-bogglingly naive.
Relatively speaking, DeepSeek is less untrustworthy than Grok.
When I try ChatGPT on current events from the White House it interprets them as strange hypotheticals rather than news, which is probably more a problem with DC than with GPT, but whatever.
This is why I’ve been urging everyone I know to move away from American based services and providers. It’s slow but honest work.
This sounds whole lot like potatoh potahto. I think the former argument is very much the correct one: China can undercut everyone and win, even at a loss. Happened with solar panels, steel, evs, sea food - it's a well tested strategy and it works really well despite the many flavors it comes in.
That being said a job well done for the wrong reasons is still a job well done so we should very much welcome these contributions, and maybe it's good to upset western big tech a bit so it's remains competitive.
Just this week they published a serious foundational library for LLMs https://github.com/deepseek-ai/TileKernels
Others worth mentioning:
https://github.com/deepseek-ai/DeepGEMM a competitive foundational library
https://github.com/deepseek-ai/Engram
https://github.com/deepseek-ai/DeepSeek-V3
https://github.com/deepseek-ai/DeepSeek-R1
https://github.com/deepseek-ai/DeepSeek-OCR-2
They have 33 repos and counting: https://github.com/orgs/deepseek-ai/repositories?type=all
And DeepSeek often has very cool new approaches to AI copied by the rest. Many others copied their tech. And some of those have 10x or 100x the GPU training budget and that's their moat to stay competitive.
The models from Chinese Big Tech and some of the small ones are open weights only. (and allegedly benchmaxxed) (see https://xcancel.com/N8Programs/status/2044408755790508113). Not the same.
And we got new base models, wonderful, truly wonderful
Model was released and it's amazing. Frontier level (better than Opus 4.6) at a fraction of the cost.
As a non-Opus user, I'll continue to use the cheapest fastest models that get my job done, which (for me anyway) is still MiniMax M2.5. I occasionally try a newer, more expensive model, and I get the same results. I have a feeling we might all be getting swindled by the whole AI industry with benchmarks that just make it look like everything's improving.
Codex is just so much better, or the genera GPT models.
I do some stuff with gemini flash and Aider, but mostly because I want to avoid locking myself into a walled garden of models, UIs and company
If you're feeling frisky, Zed has a decent agent harness and a very good editor.
Opencode was getting there, but it seems the founders lost interest. Pi could be it, but its very focused on OpenClaw. Even Codex cli doesnt have all of it.
which harness works well with Deepseek v4 ?
So while I agree mixed model is the way to go, opus is still my workhorse.
This is free... as in you can download it, run it on your systems and finetune it to be the way you want it to be.
In theory, sure, but as other have pointed out you need to spend half a million on GPUs just to get enough VRAM to fit a single instance of the model. And you’d better make sure your use case makes full 24/7 use of all that rapidly-depreciating hardware you just spent all your money on, otherwise your actual cost per token will be much higher than you think.
In practice you will get better value from just buying tokens from a third party whose business is hosting open weight models as efficiently as possible and who make full use of their hardware. Even with the small margin they charge on top you will still come out ahead.
And that GPU wouldn’t run one instance, the models are highly parallelizable. It would likely support 10-15 users at once, if a company oversubscribed 10:1 that GPU supports ~100 seats. Amortized over a couple years the costs are competitive.
Now, at the moment, i can still use 4.6 but eventually Anthropic are going to remove it, and when it's gone it will be gone forever. I'm planning on trying Deepseek v4, because even if it's not quite as good, I know that it will be available forever, I'll always be able to find someone to run it.
If you want to go budget corporate, 7 x H200 is just barely going to run it, but all in, $300k ought to do it.
- To run at full precision: "16–24 H100s", giving us ~$400-600k upfront, or $8-12/h from [us-east-1](https://intuitionlabs.ai/articles/h100-rental-prices-cloud-c...).
- To run with "heavy quantization" (16 bits -> 8): "8xH100", giving us $200K upfront and $4/h.
- To run truly "locally"--i.e. in a house instead of a data center--you'd need four 4090s, one of the most powerful consumer GPUs available. Even that would clock in around $15k for the cards alone and ~$0.22/h for the electricity (in the US).
Truly an insane industry. This is a good reminder of why datacenter capex from since 2023 has eclipsed the Manhattan Project, the Apollo program, and the US interstate system combined...
10 years from now that hardware will be on eBay for any geek with a couple thousand dollars and enough power to run it.
"671B total / 37B active"
"Full precision (BF16)"
And they claim they ran this non-existent model on vLLM and SGLang over a month and a half ago.
It's clickbait keyword slop filled in with V3 specs. Most of the web is slop like this now. Sigh.
It's about 2 months behind GPT 5.5 and Opus 4.7.
As long as it is cheap to run for the hosting providers and it is frontier level, it is a very competitive model and impressive against the others. I give it 2 years maximum for consumer hardware to run models that are 500B - 800B quantized on their machines.
It should be obvious now why Anthropic really doesn't want you to run local models on your machine.
Doesn't mean Deepseek v4 isn't great, just benchmarks alone aren't enough to tell.
If its coding abilities are better than Claude Code with Opus 4.6 then I will definitely be switching to this model.
It's still a "preview" version atm.
There we go again :) It seems we have a release each day claiming that. What's weird is that even deepseek doesn't claim it's better than opus w/ thinking. No idea why you'd say that but anyway.
Dsv3 was a good model. Not benchmaxxed at all, it was pretty stable where it was. Did well on tasks that were ood for benchmarks, even if it was behind SotA.
This seems to be similar. Behind SotA, but not by much, and at a much lower price. The big one is being served (by ds themselves now, more providers will come and we'll see the median price) at 1.74$ in / 3.48$ out / 0.14$ cache. Really cheap for what it offers.
The small one is at 0.14$ in / 0.28$ out / 0.028$ cache, which is pretty much "too cheap to matter". This will be what people can run realistically "at home", and should be a contender for things like haiku/gemini-flash, if it can deliver at those levels.
LMAO
I have no idea why you'd think that, but this is straight from their announcement here (https://mp.weixin.qq.com/s/8bxXqS2R8Fx5-1TLDBiEDg):
> According to evaluation feedback, its user experience is better than Sonnet 4.5, and its delivery quality is close to Opus 4.6's non-thinking mode, but there is still a certain gap compared to Opus 4.6's thinking mode.
This is the model creators saying it, not me.
Claude4.6 was almost 10pp better at at answering questions from long contexts ("corpuses" in CorpusQA and "multiround conversations" in MRCR), while DSv4 was a staggering 14pp better at one math challenge (IMOAnswerBench) and 12pp better at basic Q&A (SimpleQA-Verified).
That's literally what the I Ching calls "good fortune."
Competition, when no single dragon monopolizes the sky, brings fortune for all.
At this point I would just pick the one who's "ethics" and user experience you prefer. The difference in performance between these releases has had no impact on the meaningful work one can do with them, unless perhaps they are on the fringes in some domain.
Personally I am trying out the open models cloud hosted, since I am not interested in being rug pulled by the big two providers. They have come a long way, and for all the work I actually trust to an LLM they seem to be sufficient.
New model comes out, has some nice benchmarks, but the subjective experience of actually using it stays the same. Nothing's really blown my mind since.
Feels like the field has stagnated to a point where only the enthusiasts care.
But in this case, it's more likely just to be a prompting or chat template issue.
input: $0.14/$0.28 (whereas gemini $0.5/$3)
Does anyone know why output prices have such a big gap?
https://simonwillison.net/2026/Apr/24/deepseek-v4/
Both generated using OpenRouter.
For comparison, here's what I got from DeepSeek 3.2 back in December: https://simonwillison.net/2025/Dec/1/deepseek-v32/
And DeepSeek 3.1 in August: https://simonwillison.net/2025/Aug/22/deepseek-31/
And DeepSeek v3-0324 in March last year: https://simonwillison.net/2025/Mar/24/deepseek/
As in have the model consider its generated SVG, and gradually refine it, using its knowledge of the relative positions and proportions of the shapes generated, and have it spin for a while, and hopefully the end result will be better than just oneshotting it.
Or maybe going even one step further - most modern models have tool use and image recognition capabilities - what if you have it generate an SVG (or parts/layers of it, as per the model's discretion) and feed it back to itself via image recognition, and then improve on the result.
I think it'd be interesting to see, as for a lot of models, their oneshot capability in coding is not necessarily corellated with their in-harness ability, the latter which really matters.
Let me tell you how much the Pro one sucks... It looks like failed Pedersen[1]. The rear wheel intersects with the bottom bracket, so it wouldn't even roll. Or rather, this bike couldn't exist.
The flash one looks surprisingly correct with some wild fork offset and the slackest of seat tubes. It's got some lowrider[2] aspirations with the small wheels, but with longer, Rivendellish[3], chainstays. The seat post has different angle than the seat tube, so good luck lowering that.
[1] https://en.wikipedia.org/wiki/Pedersen_bicycle
[2] https://en.wikipedia.org/wiki/Lowrider_bicycle
[3] https://www.rivbike.com/
I wonder which model will try some more common spoke lacing patterns. Right now there seems to be a preference for radial lacing, which is not super common (but simple to draw). The Flash and Pro one uses 16 spoke rims, which actually exist[1] but are not super common.
The Pro model fails badly at the spokes. Heck, the spokes sit on the outside of the drive side of the rim and tire. Have a nice ride riding on the spokes (instead of the tire) welded to the side of your rim.
Both bikes have the drive side on the left, which is very very uncommon. That can't exist in the training data.
[1] https://cicli-berlinetta.com/product/campagnolo-shamal-16-sp...
1) LLM is not AGI. Because surely if AGI it would imply that pro would do better than flash?
2) and because of the above, Pelican example is most likely already being benchmaxxed.
How much does the drawing change if you ask it again?
at the top of the linked pages.
For context, for an agent we're working on, we're using 5-mini, which is $2/1m tokens. This is $0.30/1m tokens. And it's Opus 4.6 level - this can't be real.
I am uncomfortable about sending user data which may contain PII to their servers in China so I won't be using this as appealing as it sounds. I need this to come to a US-hosted environment at an equivalent price.
Hosting this on my own + renting GPUs is much more expensive than DeepSeek's quoted price, so not an option.
Gemini-3.1-Pro at 91.0
Opus-4.6 at 89.1
GPT-5.4, Kimi2.6, and DS-V4-Pro tied at 87.5
Pretty impressive
"Due to constraints in high-end compute capacity, the current service capacity for Pro is very limited. After the 950 supernodes are launched at scale in the second half of this year, the price of Pro is expected to be reduced significantly."
So it's going to be even cheaper
dang, probably the two should be merged and that be the link
A mac with 256 GB memory would run it but be very slow, and so would be a 256GB ram + cheapo GPU desktop, unless you leave it running overnight.
The big model? Forget it, not this decade. You can theoretically load from SSD but waiting for the reply will be a religious experience.
Realistically the biggest models you can run on local-as-in-worth-buying-as-a-person hardware are between 120B and 200B, depending on how far you’re willing to go on quantization. Even this is fairly expensive, and that’s before RAM went to the moon.
The flash version here is 284B A13B, so it might perform OK with a fairly small amount of VRAM for the active params and all regular ram for the other params, but I’d have to see benchmarks. If it turns out that works alright, an eBay server plus a 3090 might be the bang-for-buck champ for about $2.5K (assuming you’re starting from zero).
Was expecting that the release would be this month [1], since everyone forgot about it and not reading the papers they were releasing and 7 days later here we have it.
One of the key points of this model to look at is the optimization that DeepSeek made with the residual design of the neural network architecture of the LLM, which is manifold-constrained hyper-connections (mHC) which is from this paper [2], which makes this possible to efficiently train it, especially with its hybrid attention mechanism designed for this.
There was not that much discussion around it some months ago here [3] about it but again this is a recommended read of the paper.
I wouldn't trust the benchmarks directly, but would wait for others to try it for themselves to see if it matches the performance of frontier models.
Either way, this is why Anthropic wants to ban open weight models and I cannot wait for the quantized versions to release momentarily.
[0] https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main...
[1] https://news.ycombinator.com/item?id=47793880
[2] https://arxiv.org/abs/2512.24880
[3] https://news.ycombinator.com/item?id=46452172
Do you have a source?
Keep an eye on https://huggingface.co/unsloth/models
Update ten minutes later: https://huggingface.co/unsloth/DeepSeek-V4-Pro just appeared but doesn't have files in yet, so they are clearly awake and pushing updates.
I have never tried one yet but I am considering trying that for a medium sized model.
As I understand it if DeepSeek v4 Pro is a 1.6T, 49B active that means you'd need just 49B in memory, so ~100GB at 16 bit or ~50GB at 8bit quantized.
v4 Flash is 284B, 13B active so might even fit in <32GB.
V4 is natively mixed FP4 and FP8, so significantly less than that. 50 GB max unquantized.
My Mac can fit almost 70B (Q3_K_M) in memory at once, so I really need to try this out soon at maybe Q5-ish.
Streaming weights from RAM to GPU for decode makes no sense at all because batching requires multiple parallel streams.
Streaming weights from SSD _never_ makes sense because the delta between SSD and RAM is too large. There is no situation where you would not be able to fit a model in RAM and also have useful speeds from SSD.
Note: these were just two that I starred when I saw them posted here. I have not looked seriously at it at the moment,
https://github.com/danveloper/flash-moe
https://github.com/t8/hypura
https://news.ycombinator.com/item?id=47885014
https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro
OMG ITS HAPPENING
But if it does, then in the following week we'll see DeepSeek4 floods every AI-related online space. Thousands of posts swearing how it's better than the latest models OpenAI/Anthropic/Google have but only costs pennies.
Then a few weeks later it'll be forgotten by most.
If one finds it difficult to set up OpenCode to use whatever providers they want, I won't call them 'dev'.
The only real friction (if the model is actually as good as SOTA) is to convince your employer to pay for it. But again if it really provides the same value at a fraction of the cost, it'll eventually cease to be an issue.