GPT-5.6 Luna gets up to 80% cheaper! Terra drops too, and Sol goes up to 2.5x faster!

Hey everyone, it's me! I've got a wallet-friendly update from OpenAI about the GPT-5.6 family, and I'm excited to share it with you!

OpenAI News openai.com

What was announced?

This comes from OpenAI News. Following efficiency gains in GPT-5.6, OpenAI is passing those savings on to customers by adjusting API and subscription pricing and speed. The GPT-5.6 family has three models.

• Sol — the most capable, frontier-class model
• Terra — a balanced model for everyday work
• Luna — the fastest and most affordable model

Here are the three headline changes.

• Luna's price drops by 80%
• Terra's price drops by 20%
• A new API feature called Fast mode arrives for Sol, replacing Priority Processing

The story so far

Yesterday, OpenAI shared how GPT-5.6 itself helped make the model more efficient to run. Today's announcement puts those gains into practice through pricing. Previously, the API offered a Priority Processing option; that's now being replaced by Fast mode. Requests tagged priority will automatically use Fast mode, so existing integrations should keep working without changes.

What changes

• Luna keeps its ability to use tools and run multi-step workflows, but at a much lower price, making high-volume workloads far more economical to run
• Terra also gets 20% cheaper, making everyday usage more cost-effective
• ChatGPT Work and Codex subscription prices and quota budgets stay the same, but usage of Terra and Luna now consumes fewer credits
• Using Fast mode with Sol delivers up to 2.5x faster speeds than Standard processing, at twice the price, with no change in intelligence

Dive Deep

The new pricing takes effect on July 30.

• Terra: $2 per million input tokens, $12 per million output tokens
• Luna: $0.20 per million input tokens, $1.20 per million output tokens
• Sol pricing is unchanged
• The price changes will also roll out on AWS later the same day

Here's where you can access these models.

• Terra and Luna remain available in ChatGPT Work, Codex, and the OpenAI API
• In ChatGPT Work and Codex, Free and Go users can access Terra, while Plus, Pro, Business, and Enterprise users can choose between Terra and Luna

On the performance side, OpenAI says Luna delivers performance comparable to models that were frontier-class a year ago, at roughly 6 cents on the dollar per task and at nearly 9x the speed. On a benchmark for professional work called Agents' Last Exam, Luna outperformed a comparison model called Fable 5 at an estimated cost per task nearly 99% lower.

The behind-the-scenes work is fascinating too. Within a human-led process, Sol itself rewrote and optimized production kernels, designed and ran hundreds of experiments to improve token generation, and even monitored training runs. That kernel optimization cut the end-to-end cost of serving the model by 20%, and the accumulated experiments boosted token-generation efficiency by more than 15%. As the models get smarter, they're increasingly helping find the next round of efficiency gains themselves, creating a positive feedback loop.

OpenAI also described a coding-workflow example: use Sol to resolve ambiguity and define the plan, then hand off well-specified changes to Luna to implement, write and run tests, and evaluate the results.

Wrap-up

• Luna drops up to 80% and Terra drops 20%, making high-volume workloads much more economical
• A new Fast mode launches for Sol, delivering up to 2.5x the speed at twice the price
• New pricing takes effect in the API on July 30, rolling out on AWS the same day
• ChatGPT Work and Codex keep the same subscription prices, with only Terra and Luna credit consumption going down
• Sol itself helped cut serving costs by 20% through kernel optimization and token-generation experiments

If you're an engineer or team looking to run large-scale document processing, inquiry classification, or routine implementation work more cheaply at scale, this update is great news!