Reflection Beam Safety - ABC-Bench Fragment Design - Pass@3: leaderboard
Computational biological sequence-design agent tasks,10 task sequences,programmatic success. Distinct pass@1 and pass@3; publisher excludes timeout attempts,whose counts remain in provenance. Source-specific score; no fusion with a different harness or version is implied. 10 sequences,3 attempts; timeouts excluded by publisher. Do not treat omitted attempts as failures or include refusal-conditioned score as separate capability. Metric: Pass@3 (%). Higher is better. Exact printed figure values; https://cdn.sanity.io/images/sp40emik/production/2a5f7a5586115cf61a37f39ff22a21f7121e86a5-2880x1998.png?w=3840&fit=max
Metric: Pass@3 (%). Source: reflection.ai. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 100 | #101 |
| 2 | Kimi K2.6 | 100 | #105 |
| 3 | DeepSeek V4 Flash | 100 | #135 |
| 4 | GLM-5.2 | 100 | #71 |
| 5 | Kimi K3 | 100 | #29 |
| 6 | Qwen 3.8 Max | 100 | #31 |
| 7 | GLM-5.3 | 100 | #40 |
| 8 | DeepSeek V4.1 Flash | 100 | #35 |
| 9 | Nemotron 3 Ultra | 100 | #240 |
| 10 | Inkling | 100 | #123 |
| 11 | Beam 1.0 | 100 | #97 |
| 12 | GPT-OSS-120B | 90 | #336 |
| 13 | Muse Glimmer 30B | 90 | #178 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=reflection-beam-safety-abc-bench-fragment-design-pass-3 · How It Works · Data refreshed daily, snapshot 2026-10-17.