GPT-6.1 Sol nearly matches Astra at a quarter of the price on RareBench
danielmckinn0n · x · 2026-10-02
On GamowLabs RareBench, GPT-6.1 Sol nearly matched Astra at roughly a quarter of the price, landing solidly on the cost Pareto frontier, while Sonnet 5.5 significantly outperformed Astra at a fraction of Opus's cost. Author is still waiting on Google DeepMind's early access approval for the missing eval. The takeaway: intelligence per dollar is improving at an astounding rate.
Related event: GPT-6.1 Sol benchmarks: near-Astra performance at a fraction of the cost(21 posts)→
More from Models
- Grok 4.7 rolls out in the Grok app for chat and research after long wait — mark_k · 2026-10-02
- OpenAI's GPT-6 Astra is shockingly good at controlling robots — binarybits · 2026-10-02
- Diffusion will be everywhere: why text diffusion models may replace autoregressive LLM inference — akbirthko · 2026-10-02
- Gemini 4 isn't even out yet — but Google's TPU advantage is being underappreciated — haider1 · 2026-10-02
- Designer: Opus 5.5 'miles better' than Astra for design; Sonnet 5.5 fast and frugal — tomjohndesign · 2026-10-02
- Polymarket Puts Just 10% Odds on OpenAI Fully Pausing AI Training This Year — Polymarket · 2026-10-02