105 hidden bugs tested: DeepSeek V4.1 Flash near Opus 5 at 1/28th the cost
PawelHuryn · x · 2026-09-10
PawelHuryn tested DeepSeek V4.1 Flash on two repos containing 105 hidden bugs (find-and-fix), and it lands on the Pareto frontier:
- Opus 5 (max): 27 bugs, $51.33
- Grok 4.6 (max): 27, $16.96
- DeepSeek V4.1 Flash (max): 24, just $1.80
- GPT-5.6 Luna (xhigh): 23, $2.50
- Opus 5 (high): 21, $38.77
Takeaway: at 28x cheaper than Opus 5 (max) while missing only 3 fewer bugs, DeepSeek V4.1 Flash is a very strong everyday model. More effort levels and DeepSeek V4 Pro results are being posted hourly in the thread.
More from coding & agent
- V4.1's agent-team decisions look reasonable on single-file projects, but subagents remain half-baked — teortaxesTex · 2026-09-10
- How to build 40+ second continuous talking videos with LTX 2.5 in ComfyUI — Majestic-Regret-3030 · 2026-09-10
- Redis LangCache claims 90% LLM cost cuts; KV, prefix, prompt and semantic caching explained — blaizedsouza · 2026-09-10
- exe.dev: Spin Up Agent-Embedded VMs With a Single Prompt, Says Early User — davidcrawshaw · 2026-09-10
- Tinkerer builds persistent memory and multi-agent chat experiment inside ChatGPT — epicskyes · 2026-09-10
- Paid Fable 5 to make a short film — it rented a PC, wrote, generated and edited everything — kleffew94 · 2026-09-10