Reflection AI's open-source Beam-1.0-501B enters Agent Arena top 15 labs, ranked #6 among US labs
arena · x · 2026-10-17
Reflection AI's Beam-1.0-501B-A23B has landed in Agent Arena, making it a top-15 lab overall (#13) and #6 among US labs.
- Ranks #12 among open-weight models on Agent Arena (-3.67% Net Improvement), ahead of Hy3, MiMo V2.5 Pro and MiniMax M3 — the highest-ranked US open model
- #22 among open models on Code Arena: WebDev (1438 pts); 24 of the top 25 open models come from Chinese labs, with Beam 1.0 the sole exception
- Fully open under Apache 2.0
- Wide margin of error (±3.84%), statistically close to the Qwen3.8 models just above it
The leaderboard chart shows Claude Opus 5.5 (Max) leading at +18.9%, followed by GPT 6 Astra (+14.1%) and Gemini 4 Argon (+13.2%); Beam-1.0 sits at #13 with -3.7%.
More from Models
- Model tweaks barely move MMLU but easily break generative tasks — aryaman2020 · 2026-10-17
- Claude reverse-engineers 1977 DES test vectors to a PDP-11 rand() seed — chrisrohlf · 2026-10-17
- METR reports on OpenAI and Anthropic models required lab review and approval — kevinnbass · 2026-10-17
- Claude cracks 30-year mystery: ECDSA curve seeds were just RNG output — scaling01 · 2026-10-17
- KOL teases mystery model argon, says people are 'severely underestimating' it — iruletheworldmo · 2026-10-17
- Reddit post argues Anthropic has led benchmarks for 6+ months and is 2-3 iterations ahead internally — Concern-Excellent · 2026-10-17