DeepSeek
Chinese AI lab's open-weight large language models, offered via chat app and low-cost API.
Liked for very low cost and fast output123, with quality close enough to frontier models for most coding work.456 The most common caveat is that it trails top models like Opus on very hard tasks and UI design789, and some avoid it over data-privacy concerns.101112
Why people recommend it
Low cost · 3
Commenters call it dirt cheap and say they rarely spend much on it even with heavy use.
“It's dirt cheap, blazing fast, and is more than enough for most coding work.”
“struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable”
Show 1 more
“I stick with deepseek because it's more than good enough and is super cheap.”
Good-enough quality · 3
Many find the gap to frontier models small or unnoticeable for most of their tasks.
“These chinese models are simply good enough for the vast majority of tasks I run”
“Then I switched to DeepSeek and haven’t noticed any drop in quality. It’s fast at around 300 TPS and cheap.”
Show 1 more
“now I just do planning with DeepSeek too and don't see much difference”
Speed · 3
Users describe it as fast, with one citing around 300 tokens per second.
“It's dirt cheap, blazing fast, and is more than enough for most coding work.”
“Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper”
Show 1 more
“Then I switched to DeepSeek and haven’t noticed any drop in quality. It’s fast at around 300 TPS and cheap.”
Daily driver · 3
Several use it for nearly everything and say it executes consistently.
Show 1 more
“I use Deepseek for some home stuff, and it's as you describe - fast, consistent.”
Local running · 2
Some run quantized versions on their own hardware for real work.
“I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well.”
“run a two-bit quant of deepseek on my Strix Halo”
Fewer refusals · 3
Users note it helped with tasks other models refused and is available uncensored through some services.
“I use DS straight off of DeepSeek's API.”
“I pay $10 a month for an AI Studio with an uncensored or de-censored Deepseek”
Show 1 more
Caveats and dissent
Hard tasks · 3
It falls short of Opus on UI design and can fail consistently on very hard problems.
“I use Deepseek for some home stuff, and it's as you describe - fast, consistent.”
Show 1 more
Data privacy · 3
Commenters note training on inputs, data sent to China, and employers wary of Chinese models.
“The downside is they train their models on your input, which might be a deal-breaker for many users (as it is for me).”
“very cheap large-scale code analyses and things that require mega context windows”
Show 1 more
“I use Deepseek for some home stuff, and it's as you describe - fast, consistent.”
Context quirks · 2
It sometimes wrongly believes it is running out of context, and long contexts can cause attentional drift.
“I use the direct API from deepseek using Opencode, no problem to report, works like a charm.”
“very cheap large-scale code analyses and things that require mega context windows”
Mixed results · 2
One user got nothing useful from it, and another argues review time outweighs its API savings.
“it did nothing good for me in the scheme of things”
“A single turn around in a PR costs more in human time than the difference between deepseek and fable in API costs.”
Local speed · 1
Running it locally can be slow and sometimes needs retries.
“I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well.”
Tooling jank · 1
Its harness with computer use and browser integration is described as somewhat janky.
“It's been rather awesome”
Often compared with
14 moved from Claude to DeepSeekGLM68 comments mention both
2 moved from GLM to DeepSeek
1 moved from DeepSeek to GLM
1 moved from DeepSeek to OpenCode
3 moved from Claude Opus to DeepSeek
2 moved from Kimi to DeepSeek
3 moved from Claude Code to DeepSeekPi35 comments mention both
1 moved from OpenRouter to DeepSeek
Who recommends it in their own words
Threads where it came up
- DeepSeek-v4.1-Flash-1017 10 mentions
- DeepSeek-V4-Flash Update 9 mentions
- Ask HN: What is your (AI) dev tech stack / workflow? (June 2026) 6 mentions
- DeepSeek V4 Pro 0813 5 mentions
- DeepSeek V4 Flash 0731 5 mentions
- AI Coding at Home Without Going Broke 5 mentions
Also written as DeepSeek V4 Pro, DeepSeek V4, DeepSeek-V4-Pro, DeepSeek 4. Data: deepseek.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 45
“I used deepseek and pi harness and it decompiled it via ghidra and recompiled it for wasm in under 40 mins”
“I'm a complete amateur that had DeepSeek build a PHP2Shell exploit”
“I use DeepSeek too but it’s more of a workhorse model, requires a lot more guiding”
“I pay $10 a month for an AI Studio with an uncensored or de-censored Deepseek”
“In regular work no, and I use it a lot.”
“It's dirt cheap, blazing fast, and is more than enough for most coding work.”
“struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable”
“Ultra-cheap models like Deepseek are also very nice if you need to burn billions and billions of tokens.”
“These chinese models are simply good enough for the vast majority of tasks I run”
“I am entirely satisfied when using my Opencode Go plan with open models like Deepseek Flash/Pro, Kimi and MiMo.”
“then handed off to DeepSeek or Minimax”
“plenty of companies in UK and Europe that provide serverless access to GLM, DeepSeek etc.”
“Personal accounts I run the $17 Anthropic plan + $100 Codex Pro + DeepSeek.”
“No need to spend twenty minutes pumping a few million tokens through deepseek regardless of how cheap it is to do so.”
“the running costs alone are more expensive than DeepSeek's API and the inference quality is much lower”
“I've been using it with deepseek for my homelab. Its pretty helpful.”
“Deepseek models are awesome. I use them all the time and they keep getting better.”
“Deepseek can do 90% of the job”
“DeepSeek probably could do everything I needed.”
“I'll be sticking with Deepseek v4 for sure.”
“now I just do planning with DeepSeek too and don't see much difference”
“for the really hard functions I don't use deepseek I use opus managing other opus agents”
“daily driver since V4”
“I love DS and am not looking back.”
“Works very well and feels almost free.”
“I use Opus 5.5 and swarms of about 50 deepseek models to reverse engineer old videos games.”
“I am a huge DeepSeek fan”
“I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well.”
“I gave it a kick and got mixed results.”
“I use it for literally everything.”
“the rest can be happily done with deepseek or glm flash”
“I switched to deepseek because it was giving me better coding results than GLM, and for far less cost.”
“somehow for me only deepseek works”
“DeepSeek is great if you have a plan and you let it go.”
“I use Deepseek myself.”
“Kimi K2.5, 2.7 and Deepseek v4 were among my favorites”
“I don't see any reason GLM and DeepSeek won't eventually be as good as Claude”
“I liked DeepSeek and GLM when I used them.”
“I used to always use glm or deepseek instead of Haiku”
“along with some MiMo and DeepSeek”
“Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper”
“I use Deepseek for some home stuff, and it's as you describe - fast, consistent.”
“Then I switched to DeepSeek and haven’t noticed any drop in quality. It’s fast at around 300 TPS and cheap.”
“very cheap large-scale code analyses and things that require mega context windows”
“The downside is they train their models on your input, which might be a deal-breaker for many users (as it is for me).”