OpenAI launches cheaper GPT-6 Sol and Luna models
Sol lists at $2/$10 per million tokens, half GPT-5.6 Sol’s promotional rate, and Luna at $0.10/$0.50, OpenAI says.
Share

OpenAI expanded the GPT-6 lineup with GPT-6 Sol and GPT-6 Luna — faster / cheaper siblings trained with similar methods to GPT-6 Astra, aimed at everyday and mid-tier professional work. API prices cut 50% vs GPT-5.6 promotional tiers. Live in ChatGPT Work and Codex for Plus/Pro/Business/Enterprise/Edu (gradual rollout); Free and Go get Luna in the desktop app. API: gpt-6-sol and gpt-6-luna. Not in Chat yet.
Price: Sol $2 input / $10 output per million tokens (was $4/$20). Luna $0.10 / $0.50 (was $0.20/$1.20). Astra stays the premium pick. Caching: higher default hit rates; up to 90% discount on cached input reads; dashboard + diagnostics; effort/tool toggles that preserve cache; explicit breakpoints.
What OpenAI claims they do: Professional workflows, coding, computer use, clearer collaboration style (Astra-like: less jargon, shorter answers). Highlights (attribute to OpenAI): Sol (xhigh) 33.2% on AutomationBench at ~$0.27/task vs Claude Opus 5 max at much higher cost; Agents’ Last Exam Sol max 56.4%; FrontierCode Sol up vs 5.6 Sol; DeepSWE Sol max 68.8% (near Fable 5’s 69.9% at ~80% lower cost per task claim); Luna DeepSWE max 66.6%. OSWorld: Sol xhigh ~60.5% partial vs Opus 5 medium ~60.3% at ~80% lower cost claim.
Factuality: On OpenAI’s internal factuality eval (user-flagged error conversations), Sol makes “about half as many mistakes” as its predecessor and approaches Astra-level reliability at much lower cost — company framing, not independent audit. Luna also improves; high effort said to match GPT-5.6 Sol at ~1/100th the cost.
Alignment: Builds on Astra; company says lower rates of misleading claims about coding work vs GPT-5.6 peers — see system card.
Who for: Teams that want GPT-6-family behavior without Astra bills — Codex/agent loops, business workflows, high-volume API. Replaces “stuck on 5.6 Sol/Luna promo pricing.”
Risks: Half-price is vs 5.6 promo, not a universal undercut of every peer. Factuality and bench wins are vendor-measured. ChatGPT availability may lag API. Astra still owns the top tier.
Source: OpenAI, Introducing GPT-6 Sol and Luna.
— Rowan Keel
Edition Sept. 22, 2026 · Archive shelf


