MazeBench Chart: Open-Weight Models Still Far Behind GPT-6 Astra (16% vs 1-2%)
patience_cave · x · 2026-10-25
Concluding chart of the MazeBench test series: open-weight models have made strides but still lag GPT-6 Astra badly. On gem collection, GPT-6 Astra leads with 16% completion and by far the best move efficiency; Kimi K3.1 Pro sits at 2% and DeepSeek v4.1 Flash at 1%, both needing several times more moves. Open models still have a long way to catch up in both intelligence and efficiency.
More from Models
- Leak: OpenAI quietly restarted frontier training Oct 1 with offline web; agents found Brotli RCE — mark_k · 2026-10-25
- Claude weekly usage limit hit 96% as power user begs Anthropic for relief — adonis_singh · 2026-10-25
- ChatGPT voice mode spotted using open-source Silero VAD v6 — ObnoxiouslyVivid · 2026-10-25
- Kimi K3.1's first impressions spark buzz: 'smells like Claude' — MatthewBerman · 2026-10-25
- OrcaRouter Releases Abliterated DeepSeek V4.1 Flash With 1M Context for Verified Security Researchers — brhydon · 2026-10-25
- AI is the security debt collector: exploits now found in hours — wunderwuzzi23 · 2026-10-25