GPT-6.1 Terra scores best-in-class 8% on MazeBench at half of Sol's cost
patience_cave · x · 2026-10-27
MazeBench results show GPT-6.1 Terra scoring a best-in-class 8%, nearly matching GPT-6.1 Sol (9%) at half the cost. GPT-6 Astra leads at 16%, Gemini-4-Argon hits 12%, Claude Opus 5.5 scores 9%, while Claude Fable 5.1 and Kimi K3.1 Pro trail at 2%.
Related event: GPT-6.1 Terra nears Sol at half cost but Astra leads spatial reasoning(3 posts)→
More from Models
- Fangcun AI publishes third-party safety evaluation of DeepSeek V4.1 Flash-1017 for manipulation risks — pstAsiatech · 2026-10-28
- NVIDIA open-sources a LoRA post-training workflow to localize Nemotron with synthetic personas — NVIDIAAI · 2026-10-28
- Carbon-X hits frontier-level DNA understanding: Qwen3-14B finetune scores 66.7%, near GPT-6.1 — lvwerra · 2026-10-28
- MiniMax and Fireworks to livestream on the architecture behind MiniMax M3.1 Flash — sophiamyang · 2026-10-28
- Suspected OpenAI "6.1 Luna" spotted in Arena; ex-Vercel dev defends cheap small models — brandon_galang · 2026-10-28
- Google launches Genie 4, its most realistic interactive world model with native audio — sundarpichai · 2026-10-28