FrontierCS 2 - Multi-turn Trajectory Batches: leaderboard
FrontierCS 2 selected formal runs on 48 research engineering tasks, with each source-native coding harness. Paper v1 Table 3 fixes the effort settings; the subsequently added DeepSeek 1017 row leaves effort unspecified. Superseded attempts are not selected by score. Native bounded task scores are distinct from human-reference margins and from pass/fail verdicts. Explicit failed-run zeroes remain zero; unreported scores remain missing. Metric: Task score (0–100).
Metric: Task score (0–100). Source: frontier-cs.org. Saturation forecast: Not forecast. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (Max) | 18.46 | #5 (Claude Opus 5.5) |
| 2 | GPT-6.1 Sol (xHigh) | 18.46 | #7 (GPT-6.1 Sol) |
| 3 | DeepSeek V4.1 Flash (Max) | 11.73 | #35 (DeepSeek V4.1 Flash) |
| 4 | GLM-5.3 Flash (Max) | 11.73 | #72 (GLM-5.3 Flash) |
| 5 | GLM-5.3 (Max) | 11.69 | #39 (GLM-5.3) |
| 6 | DeepSeek V4.1 Flash (1017) | 11.69 | #27 |
| 7 | Muse Spark 1.3 (High) | 11.69 | #29 (Muse Spark 1.3) |
| 8 | Muse Wonder (Max) | 11.48 | #15 (Muse Wonder) |
| 9 | Kimi K3 (High) | 11.48 | #33 (Kimi K3) |
| 10 | Gemini 4 Argon (High) | 11.31 | #8 (Gemini 4 Argon) |
| 11 | GPT-6 Astra (xHigh) | 11.29 | #6 (GPT-6 Astra) |
No result here: #9 Claude Fable 5.1, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=frontiercs-2-multi-turn-trajectory-batches · How It Works · Data refreshed daily, snapshot 2026-10-26.