Phi-Bench - Training: leaderboard
Phi-Bench evaluates agents implementing and optimizing AI infrastructure, including kernels, serving, compression and system engineering. Scores combine correctness and performance under the paper's shared harness; task formats stay separate. This series reports Phi-Bench - Training from 2609.10226v1, Sx3.T2: Infrastructure Score (%). Higher is better.
Metric: Infrastructure Score (%). Source: arxiv.org. Saturation forecast: Around September 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 56.4 | #15 |
| 2 | Qwen 3.8 Max | 43.9 | #35 |
| 3 | GPT-5.6 Sol | 39.9 | #22 |
| 4 | Kimi K3 | 35.5 | #36 |
| 5 | Claude Sonnet 5 | 34 | #67 |
| 6 | GLM-5.2 | 28.7 | #82 |
| 7 | DeepSeek V4 Pro | 27.8 | #113 |
| 8 | Qwen 3.7 Max | 27.4 | #88 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=phi-bench-training · How It Works · Data refreshed daily, snapshot 2026-11-07.