DeepSeek V4 Flash
DeepSeek's fast, low-cost open-weight large language model from its V4 series.
HN commenters like it for being cheap and fast123, and many use it as a daily driver for coding and agent work.4567 The most common caveat is that it is seen as good enough for suitable tasks rather than the strongest model891011, and results depend on the harness and inference provider.1213
Why people recommend it
Low cost · 3
Commenters repeatedly call it cheap and the best performance per dollar.
“DeepSeek 4.1 Flash dominates perf/cost for suitable tasks.”
“I'm using deepseek-flash and its dirt cheap and effective.”
Show 1 more
“the DeepSeek 4.1 models are currently the ultimate cost-effective model the rest have to match with”
Daily coding · 3
Used every day for coding, including brownfield bug fixing and Python, C++ and CUDA projects.
“I use Deepseek v4.1 Flash to make progress on multiple projects (Python, C++, CUDA) every single day.”
Show 1 more
“I recommend trying deepseek v4.1 flash from together.ai”
Speed · 3
Reported at roughly 150-200 tokens per second, with fast time to first token.
“you can spin a bunch of DeepSeek Flash agents and each will run at ~150-200t/s”
“the DeepSeek 4.1 models are currently the ultimate cost-effective model the rest have to match with”
Show 1 more
“make the interactive experience of using deepseek 4.1-flash significantly more enjoyable than haiku IMO”
Agent workhorse · 3
Handles long agent runs and works well as a cheap worker under a planner model.
“I have been using DS Flash 4.1 pretty heavily. It runs for hours and hours and is pretty cheap.”
“deepseek flash variants, work horse.”
Show 1 more
“Coding is handled by Haiku 5.5/GPT 6 Luna/DeepSeek v4.1 Flash.”
Strong results · 3
Some commenters say it rivals frontier models, citing a one-shot Mario clone and its search.
“I use this thing daily and it’s neck and neck with SOTA models.”
“Mario one-shotted by deepseek-v4.1-flsh-1017”
Show 1 more
“So yeah, me dumb, clanker smart.”
Caveats and dissent
Not top-tier · 3
Described as good enough for suitable tasks, and one commenter calls it a mess next to GPT Sol 6.1.
“I'm still using DS-flash - it beats frontier model on speed and cost keeping good enough code quality.”
“DeepSeek 4.1 Flash dominates perf/cost for suitable tasks.”
Show 1 more
“DeepSeek 4 flash though is a mess compared to GPT Sol 6.1”
Provider variance · 2
Inference quality varies between providers and configurations, and some suspect different quantization.
“I recommend trying deepseek v4.1 flash from together.ai”
“The value you get is claimed to be 80$ in DeepSeek v4.1 Flash API value.”
Personality · 1
One commenter finds it annoying in some roles and to chat with.
“I really like DS Flash too, it can just be /really/ annoying in some roles”
Local hardware · 1
Running it locally was reported on a 128GB Mac Studio, and streaming weights from SSD lowers speed.
“I use DwarfStar DeekSeek V4 Flash on my Mac Studio with 128gb of RAM and it's pretty good - decent TPS and coding ability.”
By version sentiment of people who named that release, oldest release first
Versions with at least 5 people. 37 people didn't name a version. Small samples swing a lot, so read the counts too. Compare all models →
Often compared with
2 moved from Claude Opus to DeepSeek V4 Flash
1 moved from DeepSeek V4 Flash to QwenGLM46 comments mention both
2 moved from GLM to DeepSeek V4 Flash
1 moved from DeepSeek V4 Pro to DeepSeek V4 Flash
7 moved from Claude to DeepSeek V4 FlashKimi33 comments mention both
2 moved from Kimi to DeepSeek V4 Flash
1 moved from DeepSeek V4 Flash to Kimi
1 moved from OpenRouter to DeepSeek V4 Flash
Who recommends it in their own words
From the makers not counted in the ranking
“My own: https://huggingface.co/drawthingsai/DeepSeek-V4.1-Flash/tree/main”
“Total LLM spend so far is ~$28 in DeepSeek-V4-Flash”
Threads where it came up
- Why isn't the industry freaking out about DeepSeek 4.1 Flash? 32 mentions
- DeepSeek V4 Flash 0731 19 mentions
- DeepSeek v4.1 Flash 15 mentions
- DeepSeek-v4.1-Flash-1017 14 mentions
- Ask HN: What default model do you use and why? 14 mentions
- DeepSeek-V4-Flash Update 14 mentions
Also written as DeepSeek V4.1 Flash, DeepSeek Flash, DeepSeek 4.1 Flash, DeepSeek V4 Flash 0731. Data: deepseek-v4-flash.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 47
“4xh200 seem to be enough if you cpu offload the engrams”
“all their recent stuff (e.g. [0][1]) are MIT licensed”
“We use Mistral Large 4, GLM 5.3 Flash and DeepSeek v4.1 Flash 1017 through their API”
“post-trained on DeepSeek V4.1 Flash 1017 which is important”
“I'm still using DS-flash - it beats frontier model on speed and cost keeping good enough code quality.”
“Deepseek 4.1 Flash is turning out to be a sweet spot for me; it's cheap and blazing fast”
“right now that's DeepSeek 4.1 Flash 1017”
“you can spin a bunch of DeepSeek Flash agents and each will run at ~150-200t/s”
“I use DeepSeek flash API as a driver most of the time.”
“V4.1 Flash 1017 is my daily driver cloud model.”
“DeepSeek 4.1 Flash dominates perf/cost for suitable tasks.”
“I use Deepseek v4.1 Flash to make progress on multiple projects (Python, C++, CUDA) every single day.”
“deepseek flash variants, work horse.”
“Gemini Flash 3.8, DeepSeek Flash, and Luna are still up there as favorites”
“I have been using DS Flash 4.1 pretty heavily. It runs for hours and hours and is pretty cheap.”
“I recommend trying deepseek v4.1 flash from together.ai”
“So yeah, me dumb, clanker smart.”
“Deepseek-4.1 Flash 1017”
“I'm using deepseek-flash and its dirt cheap and effective.”
“It is terribly fast! And smart. And cheap.”
“visual intelligence capabilities are better than Deepseek 4.1 Flash”
“Speaking of, wasn't this model released a few months ago?”
“Yep, 1017 is available at the normal $0.15/$0.06 pricing”
“Deepseek Flash 4.1 is good enough”
“Coding is handled by Haiku 5.5/GPT 6 Luna/DeepSeek v4.1 Flash.”
“The value you get is claimed to be 80$ in DeepSeek v4.1 Flash API value.”
“make the interactive experience of using deepseek 4.1-flash significantly more enjoyable than haiku IMO”
“Deepseek 4.1 Flash in OpenCode.”
“This is awesome”
“https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash”
“high-throughput, low-latency large-scale inference services for DeepSeek V4.1 Flash”
“I use DS-4.1-Flash a lot but find it often gets caught up in long thinking runs.”
“I really like DS Flash too, it can just be /really/ annoying in some roles”
“They alreay did”
“Compare DS 4.1 Flash [1] to GLM 5.3 Flash [2], and consider the effective price.”
“I have a $20/month OpenAI subscription that I supplement with DeepSeek V4.1 Flash”
“not as concise and direct as DeepSeek 4.1 Flash in coding”
“the DeepSeek 4.1 models are currently the ultimate cost-effective model the rest have to match with”
“Both were great”
“Which is roughly similar to deepseek 4.1 flash.”
“DeepSeek 4 flash though is a mess compared to GPT Sol 6.1”
“I use DwarfStar DeekSeek V4 Flash on my Mac Studio with 128gb of RAM and it's pretty good - decent TPS and coding ability.”
“I've really enjoyed working with DS4F in omp”
“is quite smarter than dsv4 flash”
“DS 4 Flash, my previous daily driver, which I also found quite good”
“I use Deepseek V4 flash 0731 Q8 quantisation from unsolth with reasoning turned off on a local server for quick easy tasks”
“DeepSeek V4 Flash runs well on two Sparks”