VulcanBench v3 (BenchLM): leaderboard
AI coding agents complete the v3 coding-task suite. BenchLM republishes a source-attributed pass-at-one table; the benchmark version remains distinct from earlier VulcanBench task sets. These are published results, not BenchLM ability estimates.
Metric: Pass@1 (%). Source: benchlm.ai. Saturation forecast: Not forecast. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4.5 | 89.9 | #54 |
| 2 | Claude Fable 5 | 89.5 | #20 |
| 3 | DeepSeek V4 Flash (0731) | 88.4 | #84 |
| 4 | GPT-5.6 Sol | 87 | #22 |
| 5 | Claude Opus 5 | 87 | #16 |
| 6 | GPT-5.6 Terra | 87 | #40 |
| 7 | Grok 4.6 | 87 | #29 |
| 8 | Muse Spark 1.2 | 87 | #69 |
| 9 | GPT-5.6 Luna | 85.5 | #79 |
| 10 | Qwen 3.8 27B | 82.6 | #123 |
| 11 | Qwen 3.8 Max | 81.2 | #34 |
| 12 | GLM-5.3 | 78.3 | #44 |
| 13 | Claude Haiku 4.5 | 76.2 | #287 |
| 14 | Kimi K3 | 73.7 | #35 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vulcanbench-v3-benchlm · How It Works · Data refreshed daily, snapshot 2026-11-05.