Astra's 3D Planning: Code Execution Halves Cost for a 60% Gain Over Raw Reasoning
patience_cave · x · 2026-09-14
More MazeBench details on GPT-6 Astra: with code execution it clears every introductory level easily, rarely thinking more than 5 minutes per attempt. Without tools, it averaged 9 minutes of planning per call, sometimes over 20.
Its world-model approach delivers roughly a 60% gain at 50% less cost, though spatial reasoning still struggles in longer tunnels.
Related event: GPT-6 Astra on MazeBench: Bare Model Beats GPT-5.6 With Tools(6 posts)→
More from Models
- Qwenputer: a full RISC-V CPU inside Qwen3.5-4B weights, running C in forward passes — armand_ruiz · 2026-10-31
- OpenAI blocks weapons details for teens, adds auto-ending chats fixated on violence — rohanpaul_ai · 2026-10-31
- Granite Switch 4.1.1 ships: update models like you update software — krvarshney · 2026-10-31
- Mistral Open-Sources Le Chonk: Top Open Model on FrontierCode, Cyberdefense-Tested by CrowdStrike and Cloudflare — tensorqt · 2026-10-31
- Stanford CS336 Lecture Explains Why DPO Compares Against a Reference, Not Two Current Probabilities — stanfordnlp · 2026-10-31
- ik_llama.cpp PR #2808 adds native runtime for China Telecom's Xing4.0-29B-A4B MoE with mHC and MLA — Any_Bus9205 · 2026-10-31