FrontierSWE leaderboard: which LLMs score highest
FrontierSWE (hard multi-hour software engineering) currently has scores for 16 models. The top score is 65.5% by GPT-6 Astra (max); the median model scores 29.8% and the lowest scores 4.1%.
| # | Model | Organisation | FrontierSWE score | Cheapest input $/M | Source | Captured |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Astra (max) | OpenAI | 65.5% | $10.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 2 | Claude Opus 5.5 (max) | Anthropic | 62.3% | $4.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 3 | Claude Sonnet 5.5 (max) | Anthropic | 61.9% | $2.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 4 | Claude Fable 5.1 (max) | Anthropic | 56.3% | $10.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 5 | Claude Opus 5 (max) | Anthropic | 52.0% | $5.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 6 | Claude Fable 5 (max) | Anthropic | 47.0% | $10.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 7 | GPT-5.6 Sol (max) | OpenAI | 32.2% | $2.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 8 | GLM-5.3 (max) | Z.ai (Zhipu AI) | 30.2% | $0.070 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 9 | Grok 4.7 (xhigh) | xAI | 29.5% | $2.00 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 10 | Kimi K3 (max) | Moonshot | 25.9% | $1.29 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 11 | Grok 4.6 (xhigh) | xAI | 25.3% | $1.25 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 12 | Gemini 3.7 Flash (high) | Google DeepMind | 20.3% | $0.750 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 13 | Gemini 3.8 Flash (high) | Google DeepMind | 19.6% | $0.750 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 14 | Qwen3.8 Max (xhigh) | Alibaba | 15.8% | $1.65 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 15 | Muse Spark 1.2 (xhigh) | Meta AI | 12.0% | $1.25 | Epoch AI Benchmarking Hub | 2026-10-05 |
| 16 | Inkling (xhigh) | Thinking Machines | 4.1% | $0.950 | Epoch AI Benchmarking Hub | 2026-10-05 |
Compare all benchmarks side by side on the live leaderboard →
Among the ten highest scorers, the cheapest listed API price belongs to GLM-5.3 (max) at $0.070 per million input tokens (score 30.2%).
What FrontierSWE measures
FrontierSWE measures hard multi-hour software engineering.
How to read these scores
Scores are percentages (higher is better): the share of questions or tasks the model completed correctly. Where a model is listed at several reasoning efforts, only its highest-effort row is shown. See the methodology for how rows are chosen.
Where the data comes from
Epoch AI Benchmarking Hub (CC BY 4.0). Scores were last captured 2026-10-05; the Captured column gives each row's date.