Claude Fable 5.1 vs Kimi K2.7 Code
Compare models
Claude Fable 5.1 is more worth it — 53% vs 41%
How we score worth-it? →Top agentic coding, expensive
- It's very good at writing, very good at taking direction mid-task, and very good at working on its own. And it costs less to run.— blog ↗
- Very impressed by Fable 5.1. I'm in a thread where I constantly reply to 2-3 different things and it perfectly keeps them apart. It's like having two conversations at once.— blog ↗
- Feels less hand-wringy than Fable did.— blog ↗
- In my code reviews I agree with its changes more than the previous Fable's. And I steer it a lot more, because it handles multiple requests at once.— blog ↗
- Writing is where it impressed us the most. The text it produces reads more like it was written by a person.— blog ↗
- Fable 5.1 produced a confident code-migration plan from grep counts instead of reading the code; cheaper Codex agent found 7 errors— github
- Fable tends not to perform better, just cost more.— hn
- A month ago they were the best models I had ever worked with. Today both give me errors everywhere, on simple tasks, every day.— github
- I spent a whole day fixing bugs with Fable 5.1 and Opus 5.5 and the work output was subpar. Very disappointed.— github
- The waves of instability keep getting worse over the last year. Playing cat/mouse games with professional users is a quick way to destroy trust.— github
“Top-tier agentic coding, but you'll pay for it — one full benchmark run was $1,550”
“Great for long-horizon tasks that need real autonomy”
“For the price, most of us are better off with a mid-tier model or a coding agent”
Moonshot's open-weight coding-focused agentic model (1T MoE, 32B active, 256K context) for long-horizon software engineering.
- I find it pretty good — the ¥199 plan is plenty and has much more quota than my Codex Plus; K2.7 Code is noticeably better, tool/skill calling is active, and it no longer thinks as long as 2.6 did.— v2ex ↗
- So far it's decent — it produces my report HTML pages much faster than 2.6, which used to loop its deep thinking and crash the page.— v2ex ↗
- I tested it properly and it seems a rather decent improvement — at least it uses fewer tokens for the same task, which is reason enough for me to use it over K2.6 when I need an open model.— hn ↗
- Great! It finally follows a custom tool-call format (K2.6 couldn't). The UIs it generates are pretty good — not without problems, but certainly better than other models at this price point.— hn ↗
- Quite good, the plan is enough for me, and it has way more quota than my Codex Plus. 2.7 code is noticeably better — tool calls are proactive, and it doesn't think forever like 2.6.— v2ex ↗
- The moment OpenCode Go could use Kimi K2.7 Code I started testing and it immediately faceplanted — for a simple 'pnpm create vite' task it just deleted my pnpm-workspace.yaml and even asked whether to delete package.json and node_modules.— v2ex ↗
- Honestly just terrible — 2.7 keeps dead-looping and re-reading the same file, or getting stuck on one file and freezing; comprehension is garbage and it has almost no problem-solving ability.— v2ex ↗
- Using the ¥199 plan with Claude Code, one task burned 10% of the quota — it just kept running and running without stopping.— v2ex ↗
- Kimi K2.7 Code is 53x more expensive for cached inputs, which is 95% of my costs — the real savings with competitors is on cached inputs.— hn ↗
- I don't trust these benchmarks. I used Kimi K2.7 a number of times and was disappointed — it would run in circles on things Claude does in one shot.— hn ↗
“Noticeably better at tool/skill calling and no longer over-thinks like the previous version.”
“Uses fewer tokens for the same task.”
“Finally follows a custom tool-call format and generates pretty good UIs.”
FAQ
Is Claude Fable 5.1 or Kimi K2.7 Code more worth it?
Claude Fable 5.1 is more worth it, with a 53% worth-it score vs 41% for Kimi K2.7 Code.
How much does Claude Fable 5.1 cost vs Kimi K2.7 Code?
Claude Fable 5.1: $10/$50 per 1M tokens · Kimi K2.7 Code: $0.95/$4.00 per 1M tokens.
What do users say about Claude Fable 5.1 vs Kimi K2.7 Code?
Claude Fable 5.1: Top-tier agentic coding, but you'll pay for it — one full benchmark run was $1,550. Kimi K2.7 Code: Noticeably better at tool/skill calling and no longer over-thinks like the previous version..