How does the cost of Luna compare to the other GPT-5.6 models?

Luna is OpenAI's fastest and most affordable model, priced 80% less than the cost of the flagship Sol model.

What is the GPT-5.6 model family's performance claim?

OpenAI claims the flagship Sol outperforms Claude Fable 5 on the Artificial Analysis Coding Agent Index at less than half the cost. The Terra model performs as well as GPT-5.5 on intelligence benchmarks at half the price.

How did OpenAI achieve these efficiency gains?

The gains come from optimizations across the stack, including load balancing, speculative decoding, caching, and kernel optimization. OpenAI says its Sol model helped autonomously rewrite kernels and improve the agentic harness.

Large Language ModelsOpenAI BlogJul 29, 2026Jul 30, 2026, 06:01 JST

GPT-5.6: Sol flagship matches rival at half the cost, Terra at half price GPT-5.5

Send on LINE

3 Key Points

What happened
OpenAI announced the GPT-5.6 model family, with three tiers: flagship Sol, mid-range Terra, and budget Luna. The company states Sol outperforms Claude Fable 5 on the Artificial Analysis Coding Agent Index at less than half the cost, Terra matches GPT-5.5's intelligence at half the price, and Luna is priced 80% less than Sol.
Why it matters
GPT-5.6 is designed for greater efficiency, not just capability. OpenAI says its optimizations across the model, inference, and agentic systems have cut end-to-end serving costs by 20% and boosted token-generation efficiency by over 15%, all aimed at serving 1 billion users and 2 million businesses more cost-effectively.
What to watch
The family's efficiency hinges on whether the reported benchmark gains hold in real-world use. The post highlights that optimizations are a continuous process, with Sol working autonomously to improve its own kernels and speculative decoding, suggesting a faster pace of future cost reductions.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

OpenAI's announcement signals a shift in focus from raw intelligence to cost-per-unit of intelligence. The company frames efficiency as central to scaling its services to over 1 billion users and 2 million businesses, suggesting that lowering operational costs is key to making its AI widely accessible. The introduction of a three-tier family (Sol, Terra, Luna) is a direct attempt to offer performance at various price points, with Terra being benchmarked against its own predecessor, GPT-5.5.

The technical details reveal a strategy of compounding small gains. Reductions in cost come from system-level fixes like load balancing and caching, not just model improvements. Importantly, OpenAI's own models are being used to find these optimizations, particularly in rewriting GPU kernels, which could accelerate the pace of future improvements.

The stakes revolve around whether the benchmark claims translate to real-world user satisfaction. The test is not just if Sol can beat a competitor on a specific index, but if the entire family delivers on its promise of 'more useful intelligence per dollar' in practice. Future announcements on actual API pricing and real-world performance will be the true measure of this efficiency push.

FAQ

How does the cost of Luna compare to the other GPT-5.6 models?: Luna is OpenAI's fastest and most affordable model, priced 80% less than the cost of the flagship Sol model.
What is the GPT-5.6 model family's performance claim?: OpenAI claims the flagship Sol outperforms Claude Fable 5 on the Artificial Analysis Coding Agent Index at less than half the cost. The Terra model performs as well as GPT-5.5 on intelligence benchmarks at half the price.
How did OpenAI achieve these efficiency gains?: The gains come from optimizations across the stack, including load balancing, speculative decoding, caching, and kernel optimization. OpenAI says its Sol model helped autonomously rewrite kernels and improve the agentic harness.

OpenAI BlogRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.