Reflection Beam Preview 2026-10-05 - SWE Bench Pro v2-Hard: leaderboard
Oct5 prerelease evaluation, publisher says final red-teaming ongoing and weights forthcoming. Beam identity is date-qualified and never merged with final Oct16. Comparator exact printed model labels retained; effort/precision unreported. Public split and context-management qualifiers retained. NR cells explicitly pinned missing; a populated missing cell requires review. Metric: Score (%). Higher is better. Source: https://reflection.ai/blog/introducing-beam
Metric: Score (%). Source: reflection.ai. Saturation forecast: Around January 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K3 | 88.2 | #29 |
| 2 | GLM-5.3 | 84.3 | #40 |
| 3 | Inkling | 56.9 | #123 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=reflection-beam-preview-2026-10-05-swe-bench-pro-v2-hard · How It Works · Data refreshed daily, snapshot 2026-10-17.