The GPT-6 Sol and Luna models launched today to provide a lower cost alternative to the GPT-6 Astra architecture. API prices for these new tiers are 50% lower than GPT-5.6 promotional pricing, with Sol priced at $2 per million input tokens and $10 per million output tokens. Luna costs $0.10 per million input tokens and $0.50 per million output tokens, building on an 80% cost reduction the company implemented for that model in July.
The release is rolling out in Codex and ChatGPT Work for Plus, Pro, Business, Enterprise, and Edu users, with Luna also available to Free and Go users via the desktop app. Sol matches Claude Fable 5's performance on the DeepSWE v1.1 coding benchmark at roughly 80% lower cost per task. New caching features offer 90% discounts on cached input-token reads to reduce redundant processing during agentic workflows.
Key sources
- SOURCE@openai“50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing”x.com
- SOURCE@openai“roll out today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users”x.com
- SUPPORT@mattshumer_“I still prefer Astra/Fable 5.1 (and now, likely Opus 5.5) for my day to day”x.com
- SUPPORT@kimmonismus“Sol: $2 / $10”x.com
- SUPPORT@polynoamial“80% price cut to Luna we made at the end of July”x.com
- SUPPORT@gonzalo__nunez“move $5K of monthly spend that was "stuck" on GPT 5.2 over to a frontier model”x.com
- SUPPORT@reach_vb“On AutomationBench, Sol at xhigh effort outperforms Claude Opus 5 at max effort at just 9% of its cost per task”x.com
- SUPPORT@artificialanlys“GPT-6 Sol (max) cuts its hallucination rate from 92% to 60% and GPT-6 Luna (max) from 93% to 77%”x.com