OpenAI Launches GPT-5.6: Sol, Terra, and Luna Model Family Arrives After US Government Review

OpenAI has released GPT-5.6, its new model family of Sol, Terra, and Luna tiers, after a weeks-long US government cybersecurity review cleared the models for public rollout.

Saturday July 11, 2026 Source: openai.com
TL;DR — Quick Answer

OpenAI launched the GPT-5.6 family in July 2026 after a weeks-long review by US government cybersecurity agencies cleared the models for general availability. The family has three tiers: Sol, the flagship reasoning model; Terra, a balanced everyday model; and Luna, the most cost-efficient. Sol sets new highs on benchmarks like Agents' Last Exam at 53.6 and the Coding Agent Index at 80, while Terra and Luna undercut rivals on cost. Pricing starts at $5 input and $30 output per million tokens for Sol, down to $1 and $6 for Luna.

Key Takeaways

OpenAI Launches GPT-5.6: Sol, Terra, and Luna Model Family Arrives After US Government Review — AI news article illustration

OpenAI has released the GPT-5.6 family — anchored by flagship Sol, balanced Terra, and cost-efficient Luna — clearing general availability after a weeks-long review by US government cybersecurity agencies. The July 2026 launch delivers OpenAI’s most capable and most efficient models to date, spanning ChatGPT, Codex, and the API.

The Family

The three tiers target different workloads rather than a single upgrade path:

Benchmarks

GPT-5.6 Sol posts new state-of-the-art results across the board. On Agents’ Last Exam, it scores 53.6, eclipsing Claude Fable 5 by 13.1 points — even at medium reasoning it beats Fable 5 by 11.4 points at roughly a quarter of the estimated cost. On the Artificial Analysis Coding Agent Index it hits 80, and on cybersecurity evals such as ExploitBench it reaches 73.5 percent versus 47.9 percent for GPT-5.5. Terra and Luna outperform Fable 5 at about one-sixteenth the cost.

Pricing

The family is priced aggressively per 1 million tokens:

Programmatic tool calling and predictable prompt caching, with a 30-minute minimum cache life and a 90 percent cached-input discount, cut effective enterprise costs further.

The Government Review

General availability follows a security assessment by US government cyber agencies — a step that has become standard for frontier releases since the US adopted formal review of the most capable systems. OpenAI says GPT-5.6 ships with its most robust safeguards yet, calibrated by capability, with verified defensive users able to access more through its Trusted Access programs.

What This Means

On benchmarks, price, and safety process, GPT-5.6 resets the frontier: historically strong results at fraction-of-competitor costs, wrapped in the security review the market now expects. The Sol/Terra/Luna split gives OpenAI durable upgrade paths — the number churns, but customers pick a tier and ride it forward.

Frequently Asked Questions

What are the GPT-5.6 model tiers?

GPT-5.6 ships as three tiers: Sol, the flagship with frontier reasoning and coding performance; Terra, a balanced model for everyday work; and Luna, the fastest and most affordable. All three are available in ChatGPT, Codex, and the API.

How much does GPT-5.6 cost?

Per 1 million tokens, Sol costs $5 input and $30 output, Terra costs $2.50 and $15, and Luna costs $1 and $6. Prompt caching discounts cached reads by 90 percent.

Why did OpenAI's government review delay GPT-5.6?

A weeks-long review by US government cybersecurity agencies examined the models' cyber capabilities and safeguards before general availability, part of recent US policy requiring security assessment of the most capable frontier systems.

This article is based on the official announcement from openai.com . Read the original for full technical details.

Related Articles

Back to all news