In short: GPT-5.6 Luna costs $0.20 per million input tokens after an 80 percent price reduction and is thus cheaper than Google’s Gemini 3.1 Flash-Lite and one-fifth the price of Anthropic’s Claude Haiku 4.5.
OpenAI has cut prices for its GPT-5.6 models: Terra costs 20 percent less, Luna even 80 percent less. With this, Luna is now positioned ahead of established competing models in terms of pricing.
OpenAI has implemented a comprehensive price adjustment for its GPT-5.6 model variants. The model GPT-5.6 Terra has seen a cost reduction of 20 percent, while GPT-5.6 Luna has experienced a price cut of 80 percent. After this adjustment, Luna costs $0.20 per million input tokens and $1.20 per million output tokens.
According to OpenAI’s own statements, the company used the model GPT-5.6 Sol to optimize inference efficiency for these cost savings. This included optimizing the forward pass — the computation that transforms inputs into token predictions — as well as kernel optimizations using the GPU programming languages Triton and Gluon. The model automatically identified operations that could be precomputed, avoided, or parallelized, and rewrote production kernels with the help of Codex. These measures reduced end-to-end serving costs by 20 percent.
The Luna price cut significantly shifts the competitive landscape: At $0.20 per million input tokens, Luna is now cheaper than Google’s Gemini 3.1 Flash-Lite ($0.25 input, $1.50 output). Anthropic’s Claude Haiku 4.5, currently the most affordable model from the manufacturer, costs $1.00 input and $5.00 output — Luna is thus reduced to one-fifth of the input price.
Source: simonwillison.net · Published July 31, 2026
Lumi AI News — AI-assisted curation in accordance with Article 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.3.