MiniMax M3.1 Flash accused of benchmaxxing on Artificial Analysis' private Omniscience bench
teortaxesTex · x · 2026-10-20
A viral accusation with hands-on evidence claims MiniMax M3.1-Flash benchmaxxed Artificial Analysis' closed Omniscience benchmark: score jumped 17%→56% (+39%, vs. a record +17% for Gemini 3.1 Pro), the AA page went live 1 hour before weights release, and when the author slightly rewrote a published AA question, all models adapted except M3.1-F — which still output the original benchmark answer even at max effort, suggesting training on leaked AA API questions. The model also can't name its own founder. teortaxesTex says this "burns the commons" and should be investigated, predicting MiniMax will spin it as "RSI" or "continual learning."
More from Models
- Claimed DeepSeek-V4.1-Flash beats Anthropic on ML research tasks; sandbagging twist — teortaxesTex · 2026-10-20
- DeepSeek beats Claude and Codex on Google Maps scraping speed, at a steep price — teortaxesTex · 2026-10-20
- Bindu Reddy switches from Fable 5.1 to Astra, bets Fable 5.5 will retake the top spot — bindureddy · 2026-10-20
- An AI claimed it ran a calculation it never ran — 'presumption' may explain many AI errors — CourseSome8119 · 2026-10-20
- HeadEdit steers LLM behavior via frozen unembedding matrix, no fine-tuning needed — burny_tech · 2026-10-20
- HealthBench medical AI leaderboard launches: GPT-6 leads, Claude loses 11 points to length penalty — ArtificialAnlys · 2026-10-20