Rendered at 08:03:22 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
NitpickLawyer 2 hours ago [-]
This is more exciting than k3, IMO. Dsv4 models are extremely cheap to serve. Improving their capabilities has lots of downstream effects, as it becomes "good enough" for more and more tasks.
DS was serving the pro version at extremely low prices for a long time, and they've had integrations with opencode & other providers, so they likely gathered a lot of data from real developers doing real tasks (on openrouter they were labeled as such). Now they can use those live scenarios to further post-train their models and improve them further.
Can't wait to see if distilling k3 into dsv4 brings additional improvements. Anyway, having fast cheap models getting better is great for the community. Especially since these don't "go away" on a provider's whim. Whatever capabilities they get, can be used "forever" going forward. And, at least flash can be ran "at home" with <10k in hardware, which isn't really possible / feasible with glm/k3 larger models.
dnhkng 1 hours ago [-]
Totally! This with DwarfStar delivers usable local AI (I hope!)
f311a 17 minutes ago [-]
I've been driving flash model for 90% of my tasks. It's better than pro (for unknown reasons), very cheap and fast.
I try to keep changes under 1000 lines and drive architectural decisions myself, barely notice any difference compared to frontier models. The rest 10% is to spot bugs, security problems and to investigate better architecture, which flash can also do pretty well, I just cross check it.
Faster iterations are way better for me, I hate waiting for 5-10 minutes on small changes. I tried to use recent versions Kimi and GLM, but they use too much thinking for no reason and pretty slow. I also often feed a lot of data to it, without worrying about hitting the limits: dependencies (to find bottlenecks in them), logs, performance dumps and so on.
Also, it will never complain about security guards, I've been using it to reverse engineer binaries.
regularfry 1 minutes ago [-]
It's good but (at least on openrouter) it's got an annoyingly tight output token limit. So if it does get stuck in a reasoning pit, it won't work its way out of it in time.
It's replaced the Kimi models for me though.
Goranek 16 minutes ago [-]
Kimi K3 (instead of Opus) for expensive stuff, DSV4 Flash for tasks (instead of Sonnet)?
Does this make sense?
try-working 42 seconds ago [-]
Yep. And you can use my model router to route between them in the background.
baalimago 6 minutes ago [-]
Looks like they will release an updated version of deepseek-v4-pro soon, which most likely will beat kimi k3 at both intelligence and cost (judging by the vast improvements to dsv4 flash)
geek_at 12 minutes ago [-]
it does! I'm always amazed when I use DSV4 flash for coding or server checks and after an hour of working with it my (pure api call) balance is about 30 cents
nickandbro 1 hours ago [-]
I wouldn't doubt GPT 4.6 Luna being in the top left quadrant's center on the Cost per Intelligence Index is not concerning for Liang Wenfeng. You have to remember DeepSeek v4 flash even though a bit cheaper, does not have vision abilities, which is a big draw for agentic tasks.
I admire DeepSeek's openness, but even they have been raising prices after their discounts.
minraws 56 minutes ago [-]
They haven't raised prices though the plan was to increase it with peak hour usage for V4 Pro GA release, they didn't do that, so no price increases there I believe.
As for vision yeah it sucks but Luna is also 2x input and 1.5x output for 1M context...
That's around 0.4 in/1.8 out
DSv4 is wayyy cheaper.
And it's open now you have Luna at home if you have a decent set of GPUs you can run this on 2Sparks or one very expensive Mac or just like 6-8 5090s..
minraws 51 minutes ago [-]
I know most folks can't afford it I am working on making it viable to rent shared hosting the biggest issue is data leak and prompt injection attacks with shared hosting. (Since the server owner connects to your main system via the coding agent)
I guess using a ZDR provider is good enough for now.
dudisubekti 54 minutes ago [-]
Gpt 5.6 Luna cache read is $0.02 per mtok
V4 flash cache read is $0.0028 per mtok
That's not "a bit cheaper", just saying
nickandbro 43 minutes ago [-]
That's a good point. Yeah their caching input is insane.
re-thc 23 minutes ago [-]
> I wouldn't doubt GPT 4.6 Luna being in the top left quadrant's center on the Cost per Intelligence Index is not concerning for Liang Wenfeng.
The leaked interview has him saying it doesn't matter... as much as open source doesn't matter. There's enough in it for everyone right now and they aren't after everything.
Perspective: DeepSeek doesn't have enough infrastructure to serve their target customers already.
nickandbro 16 minutes ago [-]
Can you post the link to the leaked interview? From what I understand he has been pretty tight lipped for a guy who has a larger stake worth more in his company than Dario does in Anthropic.
Rzor 6 minutes ago [-]
[delayed]
thirtygeo 30 minutes ago [-]
For both US and China models - what standard security checks and QaQc are you all doing? We're running small gamuts to test for unsolicited jailbreaks (model jailbreaks you) and incorrect records (Fake Accuracy - as Easter Egg or common thread) meaning falsified logic or information cooked in by the developers, rather than the training data speaking for itself
miyuru 15 minutes ago [-]
Judging by the openrouter leaderboard ranking for today, it looks like Dv4F us more popular than mimov2.5.
These days cost per task is more important, and SOTA models have become expensive.
ggcr 56 minutes ago [-]
Woah, a 200B model competing with GLM-5.2 and getting close to Opus 4.8. Quite impressive.
If those numbers translate well to its general capabilities, with the great caching DeepSeek has, I feel like this model will get tons of usage.
baalimago 26 minutes ago [-]
Very promising. So it will both keep the speed and reduced price, yet exceed performance of the quite sufficient deepseek-v4-pro?
Should be extending the lead in intelligence/cost index, as deepseek-v4-flash already were the most price efficient model, which now becomes even better. Although, in the deepseek APIs, the cost is leaking all information about codebases to China.
arjie 44 minutes ago [-]
Oh my goodness what an update. I need these weights. It's an incredible model for the size. The improved tool calling etc. should be able to make my harness way simpler. This runs at mega-speed on prosumer hardware (2x RTX Pro 6000).
PhilippGille 26 minutes ago [-]
The previous V4 version wasn't called “Preview” by most inference providers. For example, the OpenRouter model slug was `deepseek/deepseek-v4-flash`. So now there will be confusion when someone talks about V4 Flash or when someone offers V4 Flash inference.
Why not call it V4.1?
try-working 24 minutes ago [-]
DeepSeek themselves called it `deepseek/deepseek-v4-flash`. Pro is still like that.
egeozcan 32 minutes ago [-]
Every time I want to have fun coding something with natural language processing, I use deepseek flash. It's just incredible for the price. I have a fairly popular app with 400 users that uses DeepSeek in the background and it still didn't hit even 50 bucks of usage in a month.
wkcheng 33 minutes ago [-]
If the benchmarks are real and reflect actual use, then this is an insane model. This 300B model outperforms the previous DS4 Pro preview model (1.8T params), and it looks like it outperforms GPT 5.6 Luna too. And it's still cheaper than Luna, even with the price decrease.
Crazy.
storywatch 20 minutes ago [-]
How's their performance in English prose? We are currently searching for cost effective ways to keep story wikis up to date.
flysoft 30 minutes ago [-]
Finally have a model with usable intelligence, at a reasonable price.
Can't imagine what Pro GA would look like, considering pro preview has only 1.6t parameters.
KronisLV 50 minutes ago [-]
Wonder how good the proper version of V4 Pro will be.
I'm still considering pulling the trigger on the annual subscription of Kimi for K3 but it's sometimes slower than I'd like (at least when compared to Anthropic) even on their Vivace plan, and the token limits on the GLM Coding subscription for GLM 5.2 were too easy to hit.
Reubend 1 hours ago [-]
They're always very understated in their update descriptions. This is actually a HUGE improvement in the model's capabilities rather than just a small tweak.
kamikazechaser 22 minutes ago [-]
The flash variant is on par with Sonnet 5 on DeepSWE (54%). Big, if true.
wolttam 1 hours ago [-]
Hooray! This model makes me very optimistic about the future of local inference. The CyberGym score stands out to me.
try-working 23 minutes ago [-]
Let's see how the market reacts.
spwa4 2 hours ago [-]
In case people want to run it, it's DeepSeek-V4-Flash-284B-A13B. So it should just barely run on a single B300, and it's small enough that it'll barely run on an M5 Max too.
lukan 16 minutes ago [-]
"it'll barely run on an M5 Max "
The max version I could order now with 128 GB?
If so, the price for local inference would be 12 000 € vs 500 000 € for a B300.
NitpickLawyer 12 minutes ago [-]
There's also the 2x spark way, which should be ~8k eur? Someone down the thread reported ~60tps for 2x sparks. That's totally usable for local inference.
You can also do 2x 6kPRO in a workstation, for ~20k.
arjie 25 minutes ago [-]
Not yet, right? That's the old DS V4 preview release. We're still waiting for the weights to come out.
wolttam 2 hours ago [-]
It runs really well on 2 DGX Sparks - 60t/s
2 hours ago [-]
tosh 41 minutes ago [-]
[dead]
dnhkng 2 hours ago [-]
DeepSeek V4 Flash (Preview → 2026-07-31)
• Terminal Bench: 56.9 → 82.7 (+25.8)
• Toolathlon: 51.8 → 70.3 (+18.5)
Compared to GPT-5.6 Terra:
• Terminal Bench: Flash 82.7 vs Terra 78.4
• Toolathlon: Flash 70.3 vs Terra 53.1
• DeepSWE: Flash 54.4 vs Terra 69.6
• Agents' Last Exam: Flash 25.2 vs Terra 50.4
Trading blows with Terra, which is pretty interesting. No clear winner on these benchmarks, and wildy differeing scores. Very interesting!
Open flash model is competing against OpenAI's 'Sonnet' model at the price of GPT 3, I am really excited about this release, hopefully it holds up in real work as well
Iolaum 1 hours ago [-]
Since they did this with their own harness I m not sure it's apples to apples comparison.
dnhkng 1 hours ago [-]
It will be fair if they release the harness though. I think now the future will be paired model-harness releases, not just weight dumps.
The performance changes are so big with the right harness that is makes sense to engineer the harness and fine-tune the model to one another from the start.
NitpickLawyer 1 hours ago [-]
> not sure it's apples to apples comparison.
They're literally comparing the previous version of the same model with the new one. It's based on the same architecture, same pre-trained model, just different post-training. It doesn't get more apples to apples than this.
dnhkng 1 hours ago [-]
I think the commenter means the Flash vs Terra benchmarks.
DS was serving the pro version at extremely low prices for a long time, and they've had integrations with opencode & other providers, so they likely gathered a lot of data from real developers doing real tasks (on openrouter they were labeled as such). Now they can use those live scenarios to further post-train their models and improve them further.
Can't wait to see if distilling k3 into dsv4 brings additional improvements. Anyway, having fast cheap models getting better is great for the community. Especially since these don't "go away" on a provider's whim. Whatever capabilities they get, can be used "forever" going forward. And, at least flash can be ran "at home" with <10k in hardware, which isn't really possible / feasible with glm/k3 larger models.
I try to keep changes under 1000 lines and drive architectural decisions myself, barely notice any difference compared to frontier models. The rest 10% is to spot bugs, security problems and to investigate better architecture, which flash can also do pretty well, I just cross check it.
Faster iterations are way better for me, I hate waiting for 5-10 minutes on small changes. I tried to use recent versions Kimi and GLM, but they use too much thinking for no reason and pretty slow. I also often feed a lot of data to it, without worrying about hitting the limits: dependencies (to find bottlenecks in them), logs, performance dumps and so on.
Also, it will never complain about security guards, I've been using it to reverse engineer binaries.
It's replaced the Kimi models for me though.
Does this make sense?
I admire DeepSeek's openness, but even they have been raising prices after their discounts.
As for vision yeah it sucks but Luna is also 2x input and 1.5x output for 1M context...
That's around 0.4 in/1.8 out
DSv4 is wayyy cheaper.
And it's open now you have Luna at home if you have a decent set of GPUs you can run this on 2Sparks or one very expensive Mac or just like 6-8 5090s..
I guess using a ZDR provider is good enough for now.
V4 flash cache read is $0.0028 per mtok
That's not "a bit cheaper", just saying
The leaked interview has him saying it doesn't matter... as much as open source doesn't matter. There's enough in it for everyone right now and they aren't after everything.
Perspective: DeepSeek doesn't have enough infrastructure to serve their target customers already.
https://openrouter.ai/rankings?view=day#leaderboard-table
These days cost per task is more important, and SOTA models have become expensive.
If those numbers translate well to its general capabilities, with the great caching DeepSeek has, I feel like this model will get tons of usage.
Should be extending the lead in intelligence/cost index, as deepseek-v4-flash already were the most price efficient model, which now becomes even better. Although, in the deepseek APIs, the cost is leaking all information about codebases to China.
Why not call it V4.1?
Crazy.
I'm still considering pulling the trigger on the annual subscription of Kimi for K3 but it's sometimes slower than I'd like (at least when compared to Anthropic) even on their Vivace plan, and the token limits on the GLM Coding subscription for GLM 5.2 were too easy to hit.
The max version I could order now with 128 GB?
If so, the price for local inference would be 12 000 € vs 500 000 € for a B300.
You can also do 2x 6kPRO in a workstation, for ~20k.
• Terminal Bench: 56.9 → 82.7 (+25.8)
• Toolathlon: 51.8 → 70.3 (+18.5)
Compared to GPT-5.6 Terra:
• Terminal Bench: Flash 82.7 vs Terra 78.4
• Toolathlon: Flash 70.3 vs Terra 53.1
• DeepSWE: Flash 54.4 vs Terra 69.6
• Agents' Last Exam: Flash 25.2 vs Terra 50.4
Trading blows with Terra, which is pretty interesting. No clear winner on these benchmarks, and wildy differeing scores. Very interesting!
Terra 87.4
https://openai.com/index/gpt-5-6/
The performance changes are so big with the right harness that is makes sense to engineer the harness and fine-tune the model to one another from the start.
They're literally comparing the previous version of the same model with the new one. It's based on the same architecture, same pre-trained model, just different post-training. It doesn't get more apples to apples than this.