LiveBench TypeScript: leaderboard
The TypeScript part of LiveBench's agentic coding category, where the model works inside a checkout of a real TypeScript repository and resolves a reported issue over many tool-using turns. Runs use a SWE agent scaffold with step and wall-clock limits, and the resulting patch is graded by the Multi-SWE-Bench harness. Higher is better. Source: github.com/LiveBench/LiveBench.
Metric: Score. Source: livebench.ai. Saturation forecast: Around July 2028. 71 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (Max) | 63.33 | #3 (Claude Opus 5.5) |
| 2 | DeepSeek V4.1 Flash (Max) | 60 | #36 (DeepSeek V4.1 Flash) |
| 3 | Claude Fable 5.1 (Max) | 60 | #9 (Claude Fable 5.1) |
| 4 | Muse Wonder (Max) | 60 | #15 (Muse Wonder) |
| 5 | Muse Spark 1.3 (xHigh) | 60 | #24 (Muse Spark 1.3) |
| 6 | Qwen 3.8 Max | 56.67 | #35 |
| 7 | DeepSeek-V4-Flash-Vision-Exp | 56.67 | #76 |
| 8 | Claude Sonnet 5.5 (Max) | 56.67 | #11 (Claude Sonnet 5.5) |
| 9 | Kimi K3 | 53.33 | #30 |
| 10 | Claude Fable 5 (Max) | 53.33 | #16 (Claude Fable 5) |
| 11 | Claude Opus 5.5 (xHigh) | 53.33 | #3 (Claude Opus 5.5) |
| 12 | Qwen 3.8 27B | 50 | #105 |
| 13 | GLM-5.3 | 50 | #39 |
| 14 | GPT-5.6 Sol (Max) | 50 | #19 (GPT-5.6 Sol) |
| 15 | Muse Spark 1.2 (xHigh) | 50 | #62 (Muse Spark 1.2) |
No result here: #8 Gemini 4 Argon, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=livebench-typescript · How It Works · Data refreshed daily, snapshot 2026-10-20.