OpenAI launches GPT-6 Sol and Luna: Sol beats Claude Opus 5 at 9% of the cost
whoiskatrin · x · 2026-09-23
OpenAI introduced GPT-6 Sol and Luna, bringing the advances behind Astra to faster, cheaper tiers.
- Stronger coding and computer use, improved factuality, alignment, and clearer answers
- On AutomationBench, Sol at xhigh effort outperforms Claude Opus 5 at max effort at just 9% of its per-task cost
- On OSWorld 2.0 offline, Luna at max effort beats GPT-5.6 Sol at medium effort at one-tenth the cost
- Sol makes roughly half as many mistakes as GPT-5.6 Sol on OpenAI's internal factuality eval of flagged conversations
- Improved prompt caching gives agents 90% discounts on cached input-token reads
Related event: OpenAI Launches GPT-6 Sol and Luna at Half the Price(57 posts)→
More from Models
- Veteran Engineer: Opus 5 Left Such a Bad Taste I Can't Use Opus Anymore — josh_wills · 2026-09-23
- GPT-6 Sol and Luna reportedly live on Gm, priced 12.9% below list from launch — markjeffrey · 2026-09-23
- Unverified: GPT-6 Sol and Luna reportedly live, said to be cheap and resold 12.9% below list — teortaxesTex · 2026-09-23
- Typesafe's decision model Jev drives a Go2 robot inside NVIDIA Isaac Sim via OM1 — paigeinsf · 2026-09-23
- Commenters argue stochastic sampling is essential — deterministic models may not be 'intelligent' — burny_tech · 2026-09-23
- Community prediction: Opus 5.5, GPT-6 Sol and Luna, Qwen4 all land fall 2026 — IanAndrewsDC · 2026-09-23