Join the discussion
Write your take first — we'll ask for email only when you're ready to publish.
- Hacker News
- It's not frontier, but it's far past what we had at the beginning of the year. It's very usable. I get great instruction compliance, tool calling, and with a trivial workflows flow it has very good long-running performance as well.by arjie
- Last weeks's discussion (591 points): https://news.ycombinator.com/item?id=49120299by CharlesW
- I'm not sure about all these benchmarks, I did some very simple tests (I have my own benchmarks https://upmaru.com/llm-tests) and these models fail, not sure if it's the inference provider or the model. They seem to be optimized for benchmarks more than real use cases. Do anything outside their distribution (even if it's not complex) they fail.
I Compared Deepseek V4 Flash 0731 (low) to Gemini 3.5 Flash Lite (minimal) and GPT 5.6 Luna (no reasoning) and Deepseek V4 Flash 0731 gets it wrong alot, where as Gemini and 5.6 Luna just gets it done.
by zacksiri - You're using it on low, that's why. There's a huge difference in performance from low to max effort.by theogravity
- It's really amazing to see how the gaps between the self hostable models and the closed models has been shrinking in the last 24 months.
And how this has been accelerating!!
I felt this very hard when I had to travel in the middle of nowhere in south america, with no network, and wanted to keep an LLM model on my macbook pro with 48GB of RAM. That was back in April 2026, a few months ago.
I downloaded Google Gemma 4 (google/gemma-4-26b-a4b) and - Oh boy - I was amazed by it's capacity!
I was able to use it to code simple things, ask it about nature, learn new stuff while traveling and make stories for the kids.
Was really amazing to observe and experiment this!
Seems to me there will be some good chance to run these great LLM locally on our hardware!
Amazing time to be alive
by geniium - I just don't know... This post sounds like an Ai bot.by UberFly
- The recently announced they're raising their prices 10x right?
Which would put them... exactly where everyone else is on this graph.
Edit: I seem to have misunderstood the news. I thought the magical cache read pricing was going away (0.002) and they were going to be on par with everyone else (0.02). But I have no idea.
Edit 2: Apparently, neither do they!
>We plan to raise the overall pricing for DeepSeek API services in the near future, with a significant increase expected. Please plan your usage accordingly. The specific pricing plan will be subject to official notice.
by andai - > The recently announced they're raising their prices 10x right?
No.
They sent an email to customers saying that they will raise prices "significantly".
How much that will be is speculation.
My guess is that they will just remove the 75% discount they gave when they released V4 preview. It will still be relatively cheap even at 4x the current price.
- Where does this "10x" comes from?by guilamu
- Compared to the last Deepseek V4 Flash version I've had tons of issues with it getting in infinite loops and talking to itself without executing tool calls, wasting tons of tokens
This is on Pi agent, nothing fancy at all about my prompts or use case. Anyone else experiencing this?
I've also had it randomly go from talking about Rust to talking about the electric chair, controversies about D&D rules (both irrelevant and something I've never discussed) and it's completely blind to it in future prompts even when its pointed out and referenced directly
All this said its still worth it but the agentic performance has degraded in my experience at least
by nylonstrung - Yeah, I saw the same thing - quite annoying. It can be mitigated through the prompt.by the_duke
- I've been using for work, from opencode $10/mo subscription plan, on high effort (which is better than max imo), and haven't had any issue.
When it was first available in opencode, it was kinda slow for me, I guess because everyone wanted to try the new shiny. But now it's back to being screamingly fast and Opus 4.8 level of smart, for penies.
by bel8 - using the platform.deepseek API version I haven't had this happen once in my usage (which has been exclusive since its release). Which provider are you using? I also use a pretty bare bones Pi.by hankbond
- What quantization are you using? Which infra provider?
Baseten.co's version got into a loop rather rapidly... I've since added loop detection and adjusted some other settings on the pi coding agent and have yet to notice it again. I also switched to DeepInfra ... who serves an fp4 version admittedly, but I've had no issues with it as of yet and it's the top provider on openrouter.ai volume wise.
- I think this one requires a bit of strong prompting. I am also normally a Pi user, but my experience in OpenCode with this model has been drastically better than in Pi, where it overthinks a lot and gets distracted by random things.
It might be even better in Codex or Oh My Pi according to this bench I saw earlier: https://nitter.net/composio/status/2085330847951970801
by natrys - Kimi K3 was an interesting model only a month ago, and now we're looking at the same performance for 1/20th of the price. Wild how fast this is advancing.by 542458
- Not for long, Deepseek is saying they will have a significant price jump soon. They really shouldn’t do it because they are on the cusp of capturing the scalable API market.by dyauspitr
- And now nobody seems interested in it because the price hasn't gone down
it's still $3/$15 for all providers on openrouter
because of some Kimi license
by thehamkercat - Yeah either the benchmark isn't very useful anymore or V4 Flash is a really, really good model.by whinvik
- Real question: is there anybody that is both maintaining alpha-dev capability by keeping abreast of all these daily changes, while also reserving enough time to actually work?
Seems like we've reached the event horizon of whether AI advances are worth paying attention to.
- DeepSeek has announced an upcoming "significant increase" in price, so this line may have to move to the right soon. https://api-docs.deepseek.com/quick_start/pricing/by modeless
- Why? It's open weight, there are plenty providers on open router that are serving the latest v4 flash at 0.14/0.28 $.by vb-8448
- I strongly recommend trying this for programming tasks.
It is strong (not Fable strong though) with a much better “persona” than Opus, and very different blindspots. If you flip between Claude and this you will find both catch the mistakes of the other before they get out of control.
On balance I actually prefer DeepSeek for programming now, because of the way it talks.
by mosura - This also reflects my experience and should put to bed the distillation rumours. This model feels nothing like the Claude models, including tone and blindspots.by chorizo
- Note this is the 07/31 release of DSv4 flash and not the "preview" that they put out a couple months or so ago.
I've been running this model locally for a week, and the preview version before that. This updated one feels like a whole tier up. It's very capable for debugging and analyzing documents/data I upload.
The killer feature, IMO, is the speed. On 2x RTX Pro 6000 Blackwell, its ~8k tok/s prefill and ~250 tok/s on a single stream. I saw 1000 tok/s with ~64 concurrent streams on vLLM.
That's fast enough that you can interactively chat with it without switching tabs while you wait, and its a ~300B (13B active, hence the speed) model so the responses are also very good. It's actually more convenient now for me to direct 95%+ of my day to day usage to my local model, and only use Claude Fable for really big coding tasks.
Until this model was released, I was contemplating spending even more money on hardware to run GLM5.2 (~750B) at reasonable speeds, but I no longer feel that need. This is smart enough, and I think it only gets much better for local models from here.
by ak_t - I'm getting like 25 tok/s on 2x RTX Pro 6000. This is with llama.cpp, but I had GPT tune it for me. I was under the impression vLLM was at most ~2x faster, and usually for highly parallel loads. Any tips on where I should look first for an obvious blunder?
I'm guessing tensor parallelism or similar?
by apitman - I wish I could say that it performs reasonably on my hardware. 8x Radeon AI Pro 9700XTs, and I can't get it to hit double-digit tokens per second. Neither vllm nor llama.cpp, with various combinations of quants, draft models, and parallelism methods can get it to run at a tolerable speed. I'll be sticking with StepFun 3.7-Flash for the foreseeable future :(by MrDrMcCoy
- > On 2x RTX Pro 6000 Blackwell, its ~8k tok/s prefill and ~250 tok/s on a single stream.
For reference, on a 1x B300 it's over 400 tok/s decode on a single stream.
by kouteiheika - What runtime are you using with the 2x RTX Pro 6000 Blackwell machine? I have the same setup and tried DSv4 Flash on vLLM and ran into a ton of kernel bugs that don't seem to have been fixed yet.by namr2000
- What quantization level is that? Because official endpoints are slow.by ComputerGuru