FrontierCS 2 - Pass Rate: leaderboard
FrontierCS 2 native website aggregate snapshot over 48 research engineering tasks. These values are independently published by the website, not recomputed from the separately updated formal-run snapshot. Each source-native coding harness and reviewed Table 3 effort setting is retained; DeepSeek 1017 effort is unspecified. FCI uses Kimi K2.7 as its zero anchor; the index is not a percentage or a human score. Metric: Pass rate (%).
Metric: Pass rate (%). Source: frontier-cs.org. Saturation forecast: Around April 2028. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (Max) | 58.3 | #5 (Claude Opus 5.5) |
| 2 | GPT-6 Astra (xHigh) | 43.8 | #6 (GPT-6 Astra) |
| 3 | Gemini 4 Argon (High) | 39.6 | #8 (Gemini 4 Argon) |
| 4 | GPT-6.1 Sol (xHigh) | 35.4 | #7 (GPT-6.1 Sol) |
| 5 | GLM-5.3 (Max) | 25 | #39 (GLM-5.3) |
| 6 | Muse Wonder (Max) | 20.8 | #15 (Muse Wonder) |
| 7 | Kimi K3 (High) | 20.8 | #33 (Kimi K3) |
| 8 | DeepSeek V4.1 Flash (1017) | 16.7 | #27 |
| 9 | DeepSeek V4.1 Flash (Max) | 14.6 | #35 (DeepSeek V4.1 Flash) |
| 10 | GLM-5.3 Flash (Max) | 12.5 | #72 (GLM-5.3 Flash) |
| 11 | Muse Spark 1.3 (High) | 12.5 | #29 (Muse Spark 1.3) |
No result here: #9 Claude Fable 5.1, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=frontiercs-2-pass-rate · How It Works · Data refreshed daily, snapshot 2026-10-26.