Gemini 4 Argon ranks 6th on ErdosBench, tied with Meta Muse Wonder and MiMo v2.6 Pro
RexDouglass · x · 2026-10-15
ErdosBench, a benchmark of 226 research-level math problems inspired by Erdős, has released new rankings with Gemini 4 Argon in 6th place.
From the leaderboard (ranks 4-8):
- GPT-5.6 Sol xhigh (4th, 3.12): cleanest low-quality tail, full coverage, no D/F/M rows, but clearly above the Opus 5.5 / GPT-6 Astra cluster.
- Meta Muse Wonder (5th, 3.05): strong audit precision and deep theorem-level partials, but one serious false theorem-scope claim remains.
- Gemini 4 Argon (6th, 3.04): high mathematical yield with a major output-reliability penalty — 144 schema-valid reports average 3.337 while 82 runner-fallback trajectories average only 2.509.
- Xiaomi MiMo v2.6 Pro (7th, 3.02): high precision, well-calibrated confidence.
- GLM-5.3 Flash (8th, 2.95): top GLM-family full run with full coverage and no F-grade claims.
Meta Muse Wonder, Gemini 4 Argon, and MiMo v2.6 Pro are in a near tie despite different strengths and weaknesses.
More from Models
- Google can't rule out Loss-of-Control threshold for Gemini 4 Argon, table shows — TheZvi · 2026-10-15
- Indra OCR Fact-Check: Accuracy Claims Hold, But the 17x Speed Claim Is Really ~3x — TheOyinbooke · 2026-10-15
- Frontier AI duopoly cracking as Google and Meta ship new models — danshipper · 2026-10-15
- GPT-6 Astra solves 49 of 50 unpublished physics problems at 98%, catching physicists' errors — 141_1337 · 2026-10-15
- Rejected by NeurIPS: study shows a single test-set replica beats uncontaminated irreducible error — RylanSchaeffer · 2026-10-15
- New SWE-bench Multimodal leaderboard: Claude Opus 5.5 leads at 70.21% on visual bugs — OfirPress · 2026-10-15