Skip to content
NewsModel release

GPT-5.6 Sol, Terra, and Luna Are Now Available to All Developers via the OpenAI API

· by Pondero Newsdesk

The short version

OpenAI made GPT-5.6 Sol ($5/$30 per million tokens), Terra ($2.50/$15), and Luna ($1/$6) generally available on July 9 with no waitlist. Sol is now the default model in ChatGPT.

GPT-5.6 Sol, Terra, and Luna Are Now Available to All Developers via the OpenAI API

OpenAI made all three GPT-5.6 variants available through its self-serve API on July 9 with no waitlist. The launch set concrete per-token prices for Sol, Terra, and Luna, and switched Sol in as the default model inside ChatGPT.

What happened

The three models are available in the OpenAI Responses API as well as through ChatGPT, ChatGPT Work, and Codex. Per-million-token pricing, per TechCrunch's July 9 coverage: Sol at $5 input and $30 output; Terra at $2.50 and $15; Luna at $1 and $6. All three share a 1 million token context window and support up to 128,000 output tokens, per Simon Willison's documentation of the launch.

Sol became the default model in ChatGPT on launch day. OpenAI CEO Sam Altman told CNBC that Sol is 54% more token-efficient on coding tasks compared to prior OpenAI models, per TechCrunch's report on that interview. That figure is a vendor-supplied claim and has not been verified by independent evaluators.

On the Artificial Analysis Coding Agent Index, OpenAI reported Sol scored 80, placing it 2.8 points above Anthropic's Fable 5 while using less than half the output tokens and costing roughly one-third less, per TechCrunch. SWE-Bench Pro results favor the other side: Sol scored 64.6% against Fable 5's 80%, per Simon Willison. OpenAI separately published a critique of the SWE-Bench Pro benchmark on the same day, arguing roughly 30% of its tasks contain errors.

OpenAI described Sol as the company's "strongest cybersecurity model yet," supporting defensive work including threat modeling, code review, and blue teaming.

Why it matters

For developers running Cline with their own OpenAI API key, the pricing ladder is now concrete. Luna at $1/$6 per million tokens is the cost floor for GPT-5.6 access. Sol at $5/$30 is the ceiling. Terra sits in the middle at $2.50/$15 and, per OpenAI, delivers performance close to GPT-5.5 levels at roughly half the cost of Sol.

The benchmark picture is split. Sol's Artificial Analysis Coding Agent Index lead points to efficiency gains on agentic coding workloads. Fable 5's 15-plus-percentage-point SWE-Bench Pro advantage is large enough to matter for teams running autonomous software engineering agents. Those two indexes measure different things, and third-party evaluations have not yet settled which model holds up on real-world developer tasks.

Cursor bundles model access and does not expose per-token costs directly. The API pricing now gives teams a reference floor if they are comparing direct API access against an all-inclusive coding tool subscription.

What to watch next

Third-party benchmark comparisons of Sol against Grok 4.5 and DeepSeek V4 on coding tasks are expected by mid-July. Independent evaluations from METR and Apollo on OpenAI's cybersecurity claims are also pending. Those results will determine whether Sol's Artificial Analysis lead extends across broader developer workloads or Fable 5's SWE-Bench Pro advantage holds.

Sources