Coding · Benchmark profile
Cognition's 100-task software-engineering benchmark for whether coding agents produce mergeable, production-quality pull requests, scored for correctness, tests, scope, style, and maintainability through maintainer-authored rubrics.
Data verified 27 Jul 2026 · Methodology 1.8.0
Benchmark score on FrontierCode 1.1 Main
Claude Fable 5 leads at 53.5%, followed by Claude Opus 5 (53.4%) and Claude Opus 4.8 (46.5%).
10 ranked models · higher is better · labels show rank and score
| Model | Rank | Provider | Score |
|---|---|---|---|
| Claude Fable 5 | #1 | Anthropic | 53.5% |
| Claude Opus 5 | #2 | Anthropic | 53.4% |
| Claude Opus 4.8 | #3 | Anthropic | 46.5% |
| GPT-5.5 | #4 | OpenAI | 43% |
| Claude Sonnet 5 | #5 | Anthropic | 42.7% |
| SWE-1.7 | #6 | Cognition | 42.3% |
| Claude Opus 4.7 | #7 | Anthropic | 38.5% |
| GPT-5.4 mini | #8 | OpenAI | 27% |
| Claude Opus 4.6 | #9 | Anthropic | 26.9% |
| Claude Sonnet 4.6 | #10 | Anthropic | 24.3% |
One best score per model · higher is better
| Rank | Model | Provider | License | Evidence use | Score |
|---|---|---|---|---|---|
| #1 | Claude Fable 5 claude-fable-5 | Anthropic | closed | Reference only | 53.5% |
| #2 | Claude Opus 5 claude-opus-5 | Anthropic | closed | Reference only | 53.4% |
| #3 | Claude Opus 4.8 claude-opus-4-8 | Anthropic | closed | Reference only | 46.5% |
| #4 | GPT-5.5 gpt-5.5 | OpenAI | closed | Estimated reference | 43% |
| #5 | Claude Sonnet 5 claude-sonnet-5 | Anthropic | closed | Estimated reference | 42.7% |
| #6 | SWE-1.7 swe-1-7 | Cognition | closed | Estimated reference | 42.3% |
| #7 | Claude Opus 4.7 claude-opus-4-7 | Anthropic | closed | Reference only | 38.5% |
| #8 | GPT-5.4 mini gpt-5.4-mini | OpenAI | closed | Estimated reference | 27% |
| #9 | Claude Opus 4.6 claude-opus-4-6 | Anthropic | closed | Reference only | 26.9% |
| #10 | Claude Sonnet 4.6 claude-sonnet-4-6 | Anthropic | closed | Reference only | 24.3% |
The top of this snapshot is led by Claude Fable 5 at 53.5%; third place is 7.0 points behind. The top-10 spread is 29.2 points.
About FrontierCode 1.1 Main
Cognition's 100-task software-engineering benchmark for whether coding agents produce mergeable, production-quality pull requests, scored for correctness, tests, scope, style, and maintainability through maintainer-authored rubrics. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
Cognition's 100-task software-engineering benchmark for whether coding agents produce mergeable, production-quality pull requests, scored for correctness, tests, scope, style, and maintainability through maintainer-authored rubrics.
Claude Fable 5 by Anthropic currently leads with 53.5%.
10 models in the LuminaBench cohort have a qualifying score on this benchmark.
No. This benchmark is display-only and does not enter the overall Lumina composite.
Related
