openai launched three models at once — sol is the one you'll care about
OpenAI's GPT-5.6 family — Sol, Terra, and Luna — went live on July 9, 2026 in ChatGPT, Codex, and the API. Sol is the flagship: state-of-the-art on coding with Programmatic Tool Calling that lets the model write and run in-memory JavaScript to coordinate tools. Terra is GPT-5.5 quality at half the cost. Luna is the cheap/fast option.
OpenAI shipped the GPT-5.6 family on July 9, 2026 — three models at once under the names Sol, Terra, and Luna. they're available in ChatGPT, Codex, and the API. the release had been in limited preview since June 26 for about 20 government-approved organizations, then required regulatory clearance from the U.S. Department of Commerce before the global rollout on July 9.
the three-model naming convention is new for OpenAI. previously you picked between model numbers (GPT-5, GPT-5.5) and optional suffixes (mini, turbo). now each family has three branded tiers. knowing which tier is which takes a moment: Sol = flagship, Terra = balanced, Luna = cheap/fast.
sol: where the coding story is
GPT-5.6 Sol is OpenAI's new flagship. OpenAI claims state-of-the-art results across coding, knowledge work, cybersecurity, and science — while using fewer tokens and at lower estimated cost than GPT-5.5. the specific coding capability worth knowing about is Programmatic Tool Calling.
Programmatic Tool Calling lets Sol write and execute in-memory JavaScript to coordinate its own tool calls. instead of calling one tool at a time and waiting for a result, it can call tools in parallel, use loops and conditionals, process intermediate results, and then return a final answer. this is the model taking an orchestration role that previously required you to build a harness around it.
for vibe coding workflows: the practical implication is that Sol can manage multi-step agentic tasks — run a test suite, check which tests failed, fix the failures, run again — without you manually threading each result into the next prompt. whether that holds up against Claude Code's agent loop or Cursor's background agents in real-world codebases is still an open question. benchmarks say Sol wins; day-to-day feel requires more time.
terra: the cost play
GPT-5.6 Terra is positioned as GPT-5.5-quality performance at half the cost. if you were already comfortable with GPT-5.5 and didn't need Sol's agentic features, Terra is the obvious upgrade path — same capability, lower token bill. it's the model that will likely run most ChatGPT Plus daily use, and the one API customers should benchmark first if cost is the primary variable.
luna: fast and cheap
GPT-5.6 Luna is the smallest and most cost-efficient model in the family. OpenAI hasn't published detailed specs on Luna's coding benchmarks, but the positioning is clear: high-volume, latency- sensitive tasks where you want GPT-5-class reasoning without paying Sol or Terra prices. autocomplete, quick classification, lightweight summarization.
what changed in codex
all three models are available in Codex, OpenAI's coding-focused product. Sol is the default for complex agentic tasks in Codex. the prior default was GPT-5.5. for teams running Codex in CI pipelines or for autonomous PR review, the Programmatic Tool Calling capability in Sol is worth testing — if it handles multi-step reasoning natively, you may be able to simplify the orchestration code around your Codex integration.
a note on the naming
Sol, Terra, and Luna map to sun, earth, and moon — a neat mnemonic for size: star, planet, satellite. if that helps you remember the tier order, great. OpenAI will presumably release GPT-5.7 or GPT-6.0 at some point and restart the naming exercise, but for now the three-name pattern is at least more memorable than numbered suffixes.