LiveBench Code Completion: leaderboard
LiveBench's code completion task gives a programming problem plus the first part of a correct solution and asks the model to write the rest. The joined program must pass every test for the problem to score. It began with LeetCode problems, but since the April 2025 release the questions center on real-world libraries; the board is percent of items passed, higher is better. Source: github.com/LiveBench/LiveBench.
Metric: Score. Source: livebench.ai. Saturation forecast: Around January 2027. 71 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Luna (Max) | 86.96 | #68 (GPT-5.6 Luna) |
| 2 | Claude Opus 5.5 (Max) | 86.96 | #3 (Claude Opus 5.5) |
| 3 | GPT-5.2 Codex | 86.96 | #95 |
| 4 | Claude Sonnet 5.5 (Max) | 86.96 | #11 (Claude Sonnet 5.5) |
| 5 | Claude Opus 5.5 (xHigh) | 86.96 | #3 (Claude Opus 5.5) |
| 6 | GPT-5.6 Sol (Max) | 84.78 | #19 (GPT-5.6 Sol) |
| 7 | Claude Opus 4.8 (Max) | 84.78 | #37 (Claude Opus 4.8) |
| 8 | Claude Sonnet 5.5 (xHigh) | 84.78 | #11 (Claude Sonnet 5.5) |
| 9 | Kimi K3 | 82.61 | #30 |
| 10 | GPT-5.5 (xHigh) | 82.61 | #31 (GPT-5.5) |
| 11 | Claude Opus 5 (Max) | 82.61 | #12 (Claude Opus 5) |
| 12 | DeepSeek V4.1 Flash (Max) | 82.61 | #36 (DeepSeek V4.1 Flash) |
| 13 | Claude Fable 5.1 (Max) | 82.61 | #9 (Claude Fable 5.1) |
| 14 | GPT-6.1 Sol (xHigh) | 82.61 | #7 (GPT-6.1 Sol) |
| 15 | GLM-5.2 | 80.44 | #73 |
No result here: #8 Gemini 4 Argon, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=livebench-code-completion · How It Works · Data refreshed daily, snapshot 2026-10-20.