Benchmarks · coding
FrontierSWE v2
Cognition's ultra-long-horizon software engineering benchmark: 34 tasks graded 0-100 on implementation, performance and research; mean@5 with a 20-hour budget per task in the Proximus harness at max reasoning effort.
22 models · 22 results · data as of 2026-11-06 · frontierswe.com ↗
Leaderboard
GPT-6 Astra max
Heavy line: best score at each price. Thin lines join one model's effort levels. Cost uses each model's median subscription plan, or API price with no plan data. 12 more models under All models in settings.
More from this benchmark
1 published measuresTime per task22 results · lower is better
8 more models under All models in settings.
All results
22 rows| # | Model | Effort | FrontierSWE v2 | Subscription $ / task | API $ / task | Time | Harness |
|---|---|---|---|---|---|---|---|
| 1 | GPT-6 Astra | max | 65.5% ±8.9 | $44 | $1030 | 12.1 h | proximus |
| 2 | Opus 5.5 | max | 62.3% ±9.8 | $1.50 | $99 | 14.2 h | proximus |
| 3 | Sonnet 5.5 | max | 61.9% ±10.4 | $1.67 | $110 | 16.0 h | proximus |
| 4 | Gemini 4 Carbon | max | 61.2% ±5.3 | $22 | $62 | 12.6 h | proximus |
| 5 | Fable 5.1 | max | 56.3% ±11.1 | $22 | $139 | 11.6 h | proximus |
| 6 | Gemini 4 Argon | max | 55.0% ±9.9 | $92 | $129 | 10.6 h | proximus |
| 7 | Opus 5 | max | 52.0% ±10.8 | $3.41 | $197 | 16.1 h | proximus |
| 8 | Fable 5 | max | 47.0% ±13.9 | $34 | $301 | 15.8 h | proximus |
| 9 | Haiku 5.5 | max | 43.8% ±9.1 | $0.44 | $21 | 12.4 h | proximus |
| 10 | GPT-5.6 Sol | max | 32.2% ±11.0 | $4.96 | $180 | 8.6 h | proximus |
| 11 | GLM 5.3 | max | 30.2% ±11.5 | $10 | $97 | 17.1 h | proximus |
| 12 | Grok 4.7 | max | 29.5% ±12.6 | $30 | $319 | 12.1 h | proximus |
| 13 | Kimi K3 | max | 25.9% ±11.8 | $9.97 | $110 | 18.5 h | proximus |
| 14 | Muse Spark 1.3 | max | 25.8% ±14.2 | $122 | $243 | 17.0 h | proximus |
| 15 | Grok 4.6 | max | 25.3% ±12.2 | $25 | $243 | 13.8 h | proximus |
| 16 | Gemini 3.7 Flash | max | 20.3% ±10.5 | $4.41 | $34 | 7.9 h | proximus |
| 17 | Gemini 3.8 Flash | max | 19.6% ±11.1 | $1.57 | $39 | 7.3 h | proximus |
| 18 | GLM 5.3 Flash | max | 18.1% ±9.9 | free | $11 | 16.4 h | proximus |
| 19 | Qwen 3.8 Max | max | 17.8% ±9.4 | $6.06 | $51 | 19.4 h | proximus |
| 20 | DeepSeek V4 Flash | max | 14.8% ±9.7 | $0.44 | $8.57 | 14.9 h | proximus |
| 21 | Muse Spark 1.2 | max | 12.0% ±5.8 | $14 | $28 | 4.6 h | proximus |
| 22 | Inkling | max | 4.1% ±5.3 | $0.84 | $9.15 | 1.2 h | proximus |