DeepSeek

FIRST-HAND READS

What people reported, in their own words. Every line links to where it was said.

• “I found the best is to use something like Deepseek V4 Flash -- with a fast TPS provider -- and work on the code myself.”

  POSITIVE · performance · Hacker News · 2026-09-01

• “If I use DeepSeek, I do so knowing that I have to limit to simpler tasks on smaller, well specified prompts. What the author did, letting the model do the planning, is not something DeepSeek will excel at.”

  NEGATIVE · performance · Hacker News · 2026-09-01

• “leaning heavily toward the heavily quantized deepseek. Feels like actually opus 4.5 level”

  POSITIVE · performance · Reddit · 2026-08-15

• “if I'm going to have to wait for a slow model, I'd rather wait for DeepSeek's output”

  POSITIVE · performance · Reddit · 2026-08-15

• “I can't get deepseek to draw anything even remotely looking like a human body. Faces, it does, poorly. Bodies, it just cannot do.”

  NEGATIVE · performance · Reddit · 2026-08-14

• “deepseek and glm are grinding away again, doing more progress than claude does in the 40 minutes it takes to analyse code”

  POSITIVE · performance · Hacker News · 2026-08-14

• “Tried switching to DeepSeek for 13 days. Now back to the goal”

  NEGATIVE · Reddit · 2026-08-13

• “The reason I default to DeepSeek there is prompt caching rather than the per-token price. On a long session where the same repo context gets resent, that is where the bill actually goes.”

  POSITIVE · price · Reddit · 2026-08-13

• “DeepSeek and Kimi K3 will happily do security work, and they do it pretty well”

  POSITIVE · performance · Hacker News · 2026-08-13

• “DeepSeek models seem to be generally as good as the benchmarks indicate: Not the best, but stronger across the board than any model within an order of magnitude of its price”

  POSITIVE · performance · Hacker News · 2026-08-07

• “DeepSeek is remarkably efficient at caching and their cached token rates are crazy cheap; using it with Reasonix is free real estate, like 97% cached tokens, ends up costing like 30 cents an hour to use DeepSeek V4 Pro”

  POSITIVE · price · Hacker News · 2026-07-11

• “GPT 5.6 was picking up on quality and functional issues from DeepSeek and having it resolve them cleanly”

  POSITIVE · performance · Hacker News · 2026-07-10

• “I've gotten very good results with some vibe-coded deepseek book keeping”

  POSITIVE · performance · Hacker News · 2026-07-09

• “the Deepseek v4 flash model I am using through OpenRouter is killing me on latency”

  NEGATIVE · performance · Hacker News · 2026-07-09

• “DeepSeek V4 official API gives me 8x better caching so it was 40% cheaper in my case”

  POSITIVE · price · Hacker News · 2026-07-08

• “Isn't DeepSeek half as good and 20x cheaper? "Half as good" sounds bad but for most trivial tasks I've found it more than good enough.”

  POSITIVE · price · Hacker News · 2026-07-08

• “deepseek v4 pro was beating opus 4.7 6/10 times on our internal evals”

  MIXED · performance · Hacker News · 2026-07-07

• “DeepSeek V4 Pro was beating Opus 4.7 in 6/10 of our internal evals”

  POSITIVE · performance · Hacker News · 2026-07-07

• “had deepseek v4 pro execute and it used old versions , different libraries and cut corners”

  NEGATIVE · performance · Hacker News · 2026-07-07

• “A full day of coding with Deepseek is approx $1 in tokens, and at least for my use cases the quality is equivalent to Claude.”

  POSITIVE · price · Hacker News · 2026-07-07

• “DeepSeek is consistently among the cheapest models per task in my benchmarks, while also performing quite well”

  POSITIVE · price · Hacker News · 2026-07-06

• “DS4 (a 1-bit quantization of DeepSeek V4 Flash) runs at something like 9-13 tokens/second, with a loooong time to first token. It is not a realistic interactive coding model for agentic use.”

  NEGATIVE · performance · Hacker News · 2026-07-06

• “every agent I've tried through Claude code (including deepseek and GLM) will actively try to avoid fixing pre-existing issues”

  NEGATIVE · performance · Hacker News · 2026-07-04

• “with the help of Sonnet/DeepSeek I can finally get around and weed out all the still existing fuzzing bugs”

  POSITIVE · performance · Hacker News · 2026-07-04

• “Tokens from DeepSeek are a good deal, especially with Reasonix agent (which maximizes caching, which DeepSeek is uniquely good at, and cached tokens are uniquely cheap at DeepSeek)”

  POSITIVE · price · Hacker News · 2026-07-03