OpenAI slashes GPT-5.6 Luna prices by 80%, undercutting rivals
OpenAI announced significant price reductions for its GPT-5.6 models on 30th July 2026. GPT-5.6 Terra received a 20% discount, while GPT-5.6 Luna saw an 80% price drop. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens, making it cheaper than Google's Gemini 3.1 Flash-Lite ($0.25/$1.50) and far below Anthropic's Claude Haiku 4.5 ($1/$5). The price cuts were enabled by GPT-5.6 Sol, which optimized load balancing and inference. Sol autonomously rewrote production kernels using Triton and Gluon, reducing end-to-end serving costs by 20%. Developer Simon Willison switched his agent.datasette.io demo site from Gemini 3.1 Flash-Lite to Luna. The post is a link post by Simon Willison.
Key facts
- OpenAI reduced GPT-5.6 Terra prices by 20%.
- OpenAI reduced GPT-5.6 Luna prices by 80%.
- Luna now costs $0.20/M input tokens and $1.20/M output tokens.
- Luna is cheaper than Google Gemini 3.1 Flash-Lite ($0.25/$1.50).
- Luna is 1/5th the input cost of Anthropic Claude Haiku 4.5 ($1/$5).
- GPT-5.6 Sol optimized load balancing and inference.
- Sol rewrote production kernels using Triton and Gluon.
- End-to-end serving costs reduced by 20%.
- Simon Willison switched his demo site to Luna.
- The announcement was made on 30th July 2026.
Entities
Institutions
- OpenAI
- Anthropic
- Simon Willison
- CNBC
Sources
- Simon Willison —
- Quartz —