Kimi K2.5: what developers actually say
We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on Kimi K2.5. 28 of the 41 opinions come from people who say they use it themselves.
Kimi K2.5 has 74% approval from 41 opinions by 38 authors, with 28 firsthand. Cost is the main draw at 13 positive and 1 negative, and users put it at roughly 80% to 90% of Claude Opus quality for much less money; code quality is positive at 9 to 2, while reliability is mixed at 5 to 3 and complaints cite tool-calling errors, hangs, and inconsistent output. No 90-day approval figure is available, and monthly samples are small: approval was 64% in January, 74% in February, 100% in March, and 92% in April, with only four opinions from May through October. The recent trend is therefore thin.
Pick it if
- Want strong coding value at a low price
- Mostly run small, well-scoped tasks
- Want an open-weight model for agentic coding and research
Look elsewhere if
- Depend on reliable tool calling in OpenCode
- Need consistent architecture on complex work
- Expect fast local coding on your own hardware
Strengths and weaknesses
Each row counts opinions that mentioned the topic. Bars to the right of the line are praise, bars to the left are complaints.
How opinion has moved
What developers say
Praise
Complaints
Summaries are paraphrased. Each links to the original comment.
FAQ
Is Kimi K2.5 worth it?
Most opinion is positive at 74% approval, and cost is the clearest strength with 13 positive mentions against 1 negative. Users describe it as good value for small tasks, but several say complex work needs cleanup.
Is Kimi K2.5 better than Claude Code?
The opinion data is split. One developer preferred Kimi K2.5 over Claude Code, and one preferred Claude Code instead. Complaints about Kimi K2.5 in OpenCode say it falls well short of Claude Code on architecture and consistency.
Is Kimi K2.5 good for coding?
Code quality is positive at 9 mentions to 2 negative, and 74% of all opinions are positive. The common caveat is that small tasks go well while complex work needs corrections, and tool use is the weakest area.