GPT-6.2 Sol and Astra top RWE-Bench on cost-efficiency, moving the Pareto frontier
Jsevillamol · x · 2026-11-20
Unverified benchmark figures circulating on X: GPT-6.2 Sol and GPT-6.2 Astra are claimed to be the two most cost-efficient models on RWE-Bench. Astra scores 40% vs 37% for GPT-6 Astra at 4.3x lower cost per run; Sol scores 30% vs 25% for GPT-6.1 Sol at 2.5x lower cost. Commenters read GPT-6.2 as a bet on efficiency rather than raw capability, pushing the Pareto frontier. Treat as unconfirmed.
More from Models
- Naval-war game benchmark: author claims GPT 6.2 Astra crushes Opus 5.2 — RichardsonDx · 2026-11-20
- Qwen3.8-27B-pi-adapted GGUF quantization trends on Hugging Face — harness-hub · 2026-11-20
- One neuron per language hits 99.1% language-ID accuracy, beating dedicated models — julianweisser · 2026-11-20
- Google's Genie 4 goes live on Gemini API: steerable 720p worlds via WebRTC — _philschmid · 2026-11-20
- 10 hands-on lessons on what actually happens inside an AI after you type a prompt — techartist_ · 2026-11-20
- Redwood questions Google reusing safety results from weaker model for stronger Carbon — YvesMulkers · 2026-11-20