MiniMax M3.1 Flash ranks #2 in human eval and trains another AI at far lower cost
my_cat_can_code · x · 2026-10-20
In bake.ai's RSI Arena human eval, the newly launched MiniMax M3.1 Flash took #2 at remarkably lower cost than GPT-6 Astra—and impressed by autonomously running experiments, training a model, and improving it with human feedback. Full training artifacts and checkpoints are promised in an upcoming report.
More from Models
- Google AI Overview glitch leaks raw python_interpreter tool call into search results — FlolightC · 2026-10-20
- Insider tips Kimi to show up on LMArena this week — ChrisGPT · 2026-10-20
- "Codistilling": copying model smarts via AI coding beats distilling? — _arohan_ · 2026-10-20
- "AI is brutal": user returns to Claude after 3 weeks, switches for 1% gains — latticecut · 2026-10-20
- Users slam Anthropic's bio safeguards for triggering on harmless, unrelated prompts — BlackHC · 2026-10-20
- MiniCPM-V 4.7 runs locally at 100 tok/s on just ~3 GB of memory — NielsRogge · 2026-10-20