Kimi K3 Tech Report - Kimi Code Bench 2.0 (Codex): leaderboard
Moonshot AI's in-house Kimi Code Bench 2.0 evaluates code agents on end-to-end software engineering tasks across many languages and production stacks; here models run in OpenAI's Codex harness. Kimi K3 technical report, Table 3; score in percent, higher is better. GPT-5.6 Sol had 10 refusals and GPT-5.5 3 refusals out of 80 tasks.
Metric: Score (%). Source: raw.githubusercontent.com. Saturation forecast: Around July 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 69 | #33 (GPT-5.5) |
| 2 | GPT-5.6 Sol (Max) | 64.8 | #22 (GPT-5.6 Sol) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kimi-k3-tech-report-kimi-code-bench-2-0-codex · How It Works · Data refreshed daily, snapshot 2026-11-05.