FrontierSWE V2
Code
34 ultra-long-horizon coding, scientific computing and research tasks with normalized rewards from 0 to 1, averaged and expressed as percentages. Proximus harness, 20-hour task budget and five planned trials. V2 revises task QA and scoring relative to V1; incomplete graded trial counts are retained in observation notes.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Carbon | 61.2 |
| Claude Opus 5 | 52.0 |
| Claude Haiku 5.5 | 43.8 |
| GPT-5.6 Sol | 32.2 |
| GLM-5.3 | 30.2 |
| Grok 4.7 | 29.5 |
| Kimi K3 | 25.9 |
| Muse Spark 1.3 | 25.8 |
| Grok 4.6 | 25.3 |
| Gemini 3.7 Flash | 20.3 |
| Gemini 3.8 Flash | 19.6 |
| GLM-5.3 Flash | 18.1 |
| Qwen3.8 Max (0902) | 17.8 |
| Muse Spark 1.2 | 12.0 |
| Inkling | 4.1 |