OpenAI Ships GPT-5.6 Series — Three Models, One Reshapes Pricing

🚨 AI Tool Alert — July 12, 2026

OpenAI Ships GPT-5.6 Series — Three Models, One Reshapes Pricing

Launched July 9, 2026. Three tiers:

• Sol (flagship) — $5/m input, $30/m output. 1.05M token context. Full-codebase reasoning without chunking.


• Terra (balanced) — Mid-tier pricing. General-purpose API workhorse.


• Luna (budget) — $1/m input. 82% on Terminal-Bench. Changes the cost equation entirely.

Why this matters:

Luna at $1/million with 82% Terminal-Bench means routine tasks (code linting, log analysis, basic QA, summarization) are now economically viable for smaller teams. Sol's 1.05M context window eliminates chunking orchestration for large codebases.

Real use case — Tiered PR review pipeline:

A 50-dev team routing PR reviews by complexity:

1. Luna ($1/m) — Style checks, imports, basic patterns (~60% of reviews). Cost drops 93%.
2. Terra — API contracts, migration validation, test gaps.
3. Sol ($5/m) — New modules, cross-service changes, security reviews.

Estimated savings: ~70% on model costs while maintaining review quality.

Gotchas:

• Codex + ChatGPT Work share one usage pool — heavy Codex use eats ChatGPT quota


• Update your Codex client or you won't see the new models


• Luna's 82% Terminal-Bench means it handles most code-review tasks as well as models costing 5x more

GPT-5.6: Sol ($5/$30) | Terra (mid) | Luna ($1/m)
Context: Up to 1.05M tokens (Sol)
🔗 https://openai.com/index/gpt-5-6/