FrontierCode 1.1
2026-10-13 · General mergeability
- What it measures
- General mergeability: would the maintainer actually merge this PR? Maintainer-crafted tasks, graded by unit tests, rubrics, and verifiers. Runs flagged for unfair internet use are zeroed. On this site it is a general board and appears in the Web frontend, Android, and CLI / DevOps scenarios.
- How
- Cognition's FrontierCode 1.1 (current revision). The public page defaults to the Main set; this site imports every (model × reasoning effort) row from All reasoning levels, untruncated. The # column ranks models by their best Main effort, matching the source table. Score is the weighted rubric aggregate; Pass rate, Flag rate, Cost / rollout, and Output tokens match the source columns.
- What the score means
- The percent is the mergeability Score (higher is better). Runs flagged for unfair internet use score 0. Composite treats it as a rate board: ratio to the leader.
- Caveats
- Main and Extended are separate task sets; this page keeps the default Main board only. Score (pre-flag) and Flag rate (1.0) are null in the 1.1 Main payload, so the source page omits those columns. Rows such as SWE-* without a model page keep the source label only.
- Updates
- Revision 1.1 is current; unfair-internet runs are zeroed. This page synced 130 rows (all reasoning levels for 43 models) from https://cognition.com/data/frontiercode-leaderboard/data.json key v1_1 Main on 2026-10-13, matching the page's All reasoning levels view.
Full board
130
| # | Model | Score | Pass rate % | Flag rate | Cost / rollout | Output tokens |
|---|
| 1 | Claude Opus 5.5medium | 54.6% | 59.6% | 0.0% | $0.80 | 18.5k |
| 1 | Claude Opus 5.5max | 54.4% | 60.0% | 0.8% | $6.19 | 165.8k |
| 1 | Claude Opus 5.5high | 54.0% | 58.9% | 0.0% | $1.09 | 26.1k |
| 1 | Claude Opus 5.5xhigh | 51.4% | 56.8% | 0.0% | $2.25 | 60.2k |
| 1 | Claude Opus 5.5low | 47.3% | 51.7% | 0.0% | $0.40 | 9.1k |
| 2 | Claude Fable 5xhigh | 53.5% | 58.9% | 0.3% | $13.09 | 58.6k |
| 2 | Claude Fable 5high | 52.7% | 57.7% | 0.2% | $9.48 | 43.5k |
| 2 | Claude Fable 5max | 51.6% | 56.9% | 0.2% | $19.07 | 80.8k |
| 2 | Claude Fable 5medium | 49.8% | 54.6% | 0.3% | $7.15 | 33.2k |
| 2 | Claude Fable 5low | 48.0% | 52.8% | 0.0% | $4.92 | 23.6k |
| 3 | Claude Opus 5medium | 53.4% | 58.9% | 0.6% | $4.31 | 33.6k |
| 3 | Claude Opus 5high | 48.0% | 52.9% | 0.2% | $7.24 | 55.3k |
| 3 | Claude Opus 5max | 48.0% | 53.3% | 1.0% | $11.42 | 98.3k |
| 3 | Claude Opus 5xhigh | 43.6% | 48.4% | 1.4% | $9.14 | 67.1k |
| 3 | Claude Opus 5low | 41.9% | 46.6% | 0.2% | $2.68 | 21.9k |
| 4 | GPT-6 Astramax | 53.3% | 58.8% | 0.0% | $4.59 | 30.1k |
| 4 | GPT-6 Astrahigh | 50.9% | 56.2% | 0.0% | $3.01 | 14.3k |
| 4 | GPT-6 Astraxhigh | 50.6% | 55.8% | 0.0% | $3.28 | 17.0k |
| 4 | GPT-6 Astramedium | 48.8% | 53.6% | 0.0% | $2.43 | 10.6k |
| 4 | GPT-6 Astralow | 45.3% | 50.0% | 0.0% | $1.70 | 6.8k |