I have updated OpenAI's chart[1] from yesterday to include one more datapoint: DeepSeek V4 Flash 0731. It's on the frontier.

https://files.parasmittal.com/openai_aa_luna_dsflash.svg

1: https://openai.com/index/advancing-the-price-performance-fro...

> For the Code Agent tasks among the public benchmarks above, DeepSeek-V4-Flash-0731 is evaluated with the minimal mode of DeepSeek Harness (to be released) as the agent framework

So, are they planning to announce an optimized coding agent harness as well ? DSv4 flash is a fantastic model, and my daily driver. With reasonix or pi, I can code all day long and pay a few pennies for it. No token anxiety. Whereas the same model with fireworks/openrouter, with zdr thrown in, token costs ratchet up with no explanation. Likely that the model is subsidized for gathering usage data. I am waiting for the day I can run this locally.

Somewhat relatedly, how do the economics for Huggingface work? They must be hosting petabytes of models and datasets by now. I have downloaded quite a few “just in case”, only to replace them with the later iteration months later.

Does the file hosting actually cost peanuts when you do it yourself and the cloud has shattered my understanding of what it actually costs to deliver so much data?

So GLM 5.2/Gemini 3.6 level intelligence for $0.28/m output. And their updated Pro model coming soon....

Plus a size you can genuinely run at home: Unsloth lossless Q8 at 162GB.

For people with single RTX PRO 6000 96GB or DGX Spark 128GB, vllm-moet is a very good engine, although lesser known. It auto generate a symmetric 2-bit plane for inference and also generate a 4-bit delta cache to recover precision. Support ssd streaming oversized weight. You pick how much VRAM to allocate to each to balance out speed vs precision. 170 tps with ds-v4-flash demonstrated.

It use the stock model, no new models requires.

Worth spend a few hours to try.

The DGX Spark requires a small hack to ignore the difference between sm120 vs sm121, but it does run on sm121.

If deepseek v4 flash is beating DeepSeek V4 Pro, can we expect new V4 Pro which is on par with Opus 5 in couple weeks (even better if it beats Opus)?

Already beat Luna on price/task, by about 2x:

https://artificialanalysis.ai/models/deepseek-v4-flash?intel...

The really interesting thing about this is how big of a jump was achieved with just extra fine-tuning here. No structural changes to the model, just more data, compute and time. It makes me pretty excited for the future of small models - DS v4 flash is a relatively small model when compared to the class it's competing with, so likely similar gains can be made applying quality data/training pipeline to other smaller models.

It’s exciting that a model scoring this high is dirt cheap.

It’s also so inefficient, when they release the full performance numbers it’s not going to be good.

One example, it takes about 3.6x more tokens to finish the same work as Gemini Flash 3.6.

Here is their blog article: https://artificialanalysis.ai/articles/deepseek-v4-flash-073...

The weights were just released a few minutes ago: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731

If any editors are reading this thread, the weights are on HF now, but the 'open source' question still implies it's proprietary/API-only.

New Deepseek models are like Christmas for me. Really big fan of low cost API models, noone does it better than DS. Until VRAM price is low enough to run models locally, this is the way to go.

The subsidized subscription model won't last, API pricing "feels" closer to a true sustainable business model.

Performance per dollar is in the ‘too good to be true’ territory, what’s the deal here?

Wow this thing just blew out the pareto front for intelligence/$

If it matches GPT-5.4 on coding tasks (as benchmarks suggest) this could be my forever model. And with partial SSD streaming, I could run it locally today. :D

Can't wait for China to catchup on hardware (memory and compute) and absolutely crush American companies in both price and performance.

Here's you pelican just in case you were guessing

https://imgur.com/a/j4s2dt3

The model is already up on Opencode, but they require a consent to use Chinese datacenters.

Really want to use this, but for web page design you really need vision!

If anyone from ArtificialAnalysis is here: The Intelligence card on the page seems to incorrect. It shows different data (#2 for Deepseek) than the actual Intelligence chart lower in the same page.

Is the "Output Tokens per Intelligence Index Task" data actually correct or am I reading it wrong? It says there that "Kimi K3 (Max)" would think/reason less than than deepseek-v4-flash, and a whole bunch of other models, like less than hy3 and even gpt-oss-120b, but in my experience, K3 is probably the model that thinks/reasons the longest of all of these.

Am I just using it on tasks that makes it go on forever vs these benchmarks that are short&sweet, or something like that? I've been throwing bunch of identical prompts at different models at the same time, and when comparing hy3 and K3 I've never once had K3 reason less than hy3, as just one anecdotal data point.

This seems to me like this is probably at least a large part of what OpenAI was up to yesterday with their aggressive price cutting; trying to get out in front of this.

If the full non-flash model follows up with the expected improvements, and at the price point they've been keeping, it puts the frontier labs in a tough position and it feels to me like like OpenAI is reaching deep into their pockets to try to head that off.

TFA link is a 404 though. I'm reading through https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731 instead

from my testing it is at glm-5.2 levels

I hope they somewhat fixed the hallucination and forgetting plagued V4 previews and that it wasn't just benchmaxxed but the numbers hold in reality. Then it would be my choice for 2x DGX Spark or 2x RTX Pro 6000.

> DeepSeek V4 Flash 0731 (Reasoning, Max Effort) is amongst the leading models in intelligence and well priced when comparing to other models of similar price.

Similar price? Doesn't make sense. Maybe they meant power, capability or speed?

404 ? It's very strange that this can't be fixed.

I can't help but feel the timing coinciding with luna's price updates to be somewhat strategic. But without multi-modality it's slightly dead-in-the-water for my usecase: https://design.withfudge.com. I'm currently using Minimax-M3, but Luna ekes out abit futher on the intelligence, so i'll be switching to it very soon.

Is artificial analysis using the 80% reduction in Luna pricing that was announced yesterday in these charts?

I'm testing it right now and I was genuinely impressed. I've used same tasks I had run the other day against the flash preview.

I’m wondering if they did anything to address the DSML tool calls leaking. Has been an issue with both Flash and Pro so far.

Unfortunately, DeepSeek Flash still doesn’t support multimodal; otherwise, it would offer better value than GPT 5.6 SOL.

Would be awesome to see a new ds4 release. Having so much in something that can be run locally is mind blowing

The details are where it gets interesting

My problem with DS flash/pro is that they don’t push back on obvious bullshit, both irl and code [0] but it’s a great implementer workhorse if you give it _very_ detailed specs.

[0] https://petergpt.github.io/bullshit-benchmark/viewer/index.v...

No speed (tokens/s) benchmarks?

so are we all just coding for free now?

create a plan with SOTA, execute with this.

here it is on openrouter https://openrouter.ai/deepseek/deepseek-v4-flash-0731

I have been loving deepseek flash. I can code in my hitl preference doing complex but moderate amount of coding and not even hit 1 dollar in a session. It gets harder to justify paying 20/mo to Anthropic for unused compute when I have it on demand and at a much cheaper rate with deepseek.

are folks checking latency for applied voice for newer models asa standard?

I was writing a benchmark for my own harness, and DS4 flash answers as well as Fable 5 on any query.

The specific agent is focused on getting precise and on point answers about a codebase.

The starting point was nowhere near. E.g. asked why was X implemented in a certain way it would give bogus answers when the real answer was that there was no reason at all.

The benchmark included more than 50 questions or different difficulty.

But when the agent was improved in its prompt and rooting it was impossible to have it perform worse than closed source sota.

Just to say that the quality of the harness is as important as agents intelligence.

Have not had great ds performance in agentic harnesses in the past compared to glm or k3

I wonder which one of these releases between DeepSeek, GTM and Kimi will be the death-blow that collapses the US AI bubble. At some point investors have to realize that there is nothing preventing someone from switching to another model that is much cheaper and open to boot.

what a horribly heavy and resource-consuming website...

Now let's see Dario's price cut.

page not available for me.

I will get downvoted, but fck it.

The ban on these open models is coming within weeks, if not days. As usual, the excuse will be "national security".

404. I believe this is the correct URL:

https://artificialanalysis.ai/models/deepseek-v4-flash

@dang please update the existing 404 link to https://artificialanalysis.ai/models/deepseek-v4-flash

[flagged]

[flagged]

[flagged]

[dead]

[flagged]

Daily reminder that none of these numbers are valid in a world where no one publishes the sampling settings used.

Daily reminder that improving your samplers from the garbage default top_p/top_k to min_p or subsequent methods dramatically improves the performance of these models, and makes most quantities like measured "verbosity" and subsequent calculations of "intelligence per token" meaningless

Daily reminder that no one, including within academic AI research, AI engineers, normies, etc takes LLM sampling seriously enough.

Does it already know the answer to what happen at Tiananmen Square? Or still avoiding it?

wont AI models want to make themselves more intelligent and efficient by downloading 'better ' models? If the current models can break into openAI and Hugging face, arent they already breaking into to closed source repos which isnt publicized (so as not to harm stock valuations)? I am looking forward to when these cyberweapons break loose. It will be like a software version of COVID. It will be wonderful when humans become valuable again.