MazeBench: Kimi K3.1 Pro hits 2%, first open-weights model with a nontrivial score
patience_cave · x · 2026-10-25
On the MazeBench leaderboard, Kimi K3.1 Pro scored 2% — the first open-weights model with a nontrivial score — with DeepSeek v4.1 Flash close behind at 1%. The chart shows closed models still far ahead: gpt-6-astra (Codex max) leads at 16%, gemini-4-argon at 12%, gpt-6.1-sol and claude-opus-5-5 at 9%. Overall scores remain extremely low, highlighting how hard this benchmark is for current models.
Related event: Kimi K3.1 Pro Becomes First Open-Weight Model to Score on MazeBench(3 posts)→
More from Models
- Purify Before You Align: stripping local correlations improves diffusion alignment — LucaAmb · 2026-10-25
- AI math proofs shine at cross-discipline synthesis but lag niche specialists — jessi_cata · 2026-10-25
- Claude Code sends 16,581 tokens per step yet scores equal a 100-line mini-harness — bendee983 · 2026-10-25
- Developers vent frustration at Claude's 'deluded sense of moral superiority' — arthurcolle · 2026-10-25
- vLLM matches DeepInfer on DeepSeek-V4.1-Flash across batch sizes 8–256 — vllm_project · 2026-10-25
- I measured $20 coding plans: ~33 tasks per 5h on Sonnet, only ~6 on Opus — Sufficient-Storage87 · 2026-10-25