GPT-5.6 Sol, Terra and Luna: Specs and Pricing

What GPT-5.6 Sol, Terra and Luna are

GPT-5.6 Sol is the most capable model from OpenAI, paired with two lower-cost tiers, Terra and Luna, for different uses. All three launched as a limited preview in June 2026 and are now generally available.

How Sol, Terra and Luna differ in benchmarks and cost

GPT-5.6 Sol is OpenAI's strongest model to date. The company backed the launch with evaluations showing improved agentic capabilities (an AI deciding its own steps and carrying the work through) in three areas: coding, biology, and cybersecurity. In cyber specifically, OpenAI says Sol shifts the performance-efficiency frontier for long-horizon security tasks.

Terra and Luna trade some performance for lower cost and higher speed. Terra performs on par with the previous generation, GPT-5.5, at a price well below Sol's, so it suits high-volume workloads.

Luna, the fastest and cheapest in the family, fits light, routine work. The basic split is Sol for hard problems, Terra for high volume, and Luna for light tasks.

GPT-5.6 pricing and cost, including long context rates

GPT-5.6 is billed per token, with separate rates for input and output. All three models have had at least one price cut since launch: Terra and Luna were cut on July 30, 2026, and Sol followed with its own cut on August 21, 2026. The table below shows the current official per-1M-token rates (Standard, short context).

Cached input is discounted further: $0.40 per 1M tokens for Sol. The official pricing page also lists separate rates for cache writes and for long context—requests beyond the models' standard context window are billed at higher input and output rates. The table above covers only the basic tier.

Pricing for the wider field sits in GPT-5 in ChatGPT: Pricing, Free Tier, Models and What Is Claude AI? Free Plan, Models, vs ChatGPT, which put ChatGPT plans and Claude's model line-up side by side. Both cover the subscription side that per-token API rates leave out.

Why the top-tier model's API price fell to $4 input, $20 output

On August 21, 2026, OpenAI cut Sol's API price from $5.00/$30.00 to $4.00/$20.00 per 1M tokens (input/output), with cached input falling from $0.50 to $0.40. OpenAI describes this as promotional pricing, in effect for at least three months. Terra and Luna were not part of this cut; both have held their July 30, 2026 rates since.

GPT-5.6 benchmark performance against GPT-5.5 and Claude

OpenAI shared benchmark results across three areas: coding, biology, and cybersecurity. Sol posted the top score on Terminal-Bench 2.1, the benchmark for multi-step terminal tasks that need planning and tool use. That backs up its strength in hard coding work.

Sol also gained outside coding. On the biology benchmark GeneBench v1, which covers demanding genomics and quantitative-biology problems over many steps, Sol beat GPT-5.5 while spending fewer tokens to get there.

On the cybersecurity benchmark ExploitBench, it produced a comparable result while holding output tokens to about one-third. Doing as well or better on fewer tokens is a recurring theme of this release. OpenAI has not published a head-to-head benchmark against Claude, so any Sol-versus-Claude comparison rests on pricing and context window, not task performance.

Cyber capability, context window and the safety threshold

On the cyber side, OpenAI says Sol does not reach the top "Cyber Critical" threshold in its Preparedness Framework. In evaluations involving Chromium and Firefox, it identified real bugs and the building blocks used to assemble an exploit (attack code that abuses a vulnerability to make software behave in unintended ways).

Under the conditions tested, it did not autonomously produce a fully working end-to-end exploit. Sol's context window is 1,050,000 tokens, with a knowledge cutoff of February 16, 2026. OpenAI also says Sol ships with its strongest safety stack yet, hardened over several weeks of pressure-testing.

GPT-5.6 availability: release date, ChatGPT and Codex access

GPT-5.6 is no longer limited to a small group of partners: Sol, Terra and Luna reached general availability across ChatGPT, Codex and the API on July 9, 2026. That ended the limited preview that began June 26, 2026.

Anyone using those products, not just the trusted partners from the preview, can now use all three models. Plan-by-plan access and usage limits are laid out in GPT-5.6 availability by plan: Sol, Terra and Luna.

Codex and API access, and where GPT-6 Astra fits

Access runs through the same three surfaces as before general availability: ChatGPT, Codex, and the API, with no separate application process left to clear. On September 3, 2026, OpenAI introduced GPT-6 Astra as a new, more capable flagship, positioning it ahead of all three GPT-5.6 models. OpenAI then announced GPT-6 Sol and GPT-6 Luna at half the API price of their GPT-5.6 counterparts: $2 input and $10 output per million tokens for Sol, and $0.10 and $0.50 for Luna.

Neither Sol's model page nor its safety documentation carries a deprecated or legacy notice, though, so all three GPT-5.6 models remain available to use today. OpenAI's cyber-defense models are a separate line, covered in OpenAI Daybreak: GPT-5.5-Cyber and Codex Security.

Summary: Sol for hard work, Terra for volume, Luna for light tasks

GPT-5.6 lines up three models—the top-end Sol, the cheaper Terra, and the fastest, cheapest Luna—one for hard problems, one for high volume, and one for light tasks. All three are generally available across ChatGPT, Codex, and the API, and Sol's API price was cut on August 21, 2026, to $4.00 input / $20.00 output per 1M tokens.

On benchmarks, Sol holds the top score on Terminal-Bench 2.1 without crossing the Cyber Critical safety threshold. OpenAI has since introduced GPT-6 Astra as a new flagship, but all three GPT-5.6 models remain available, so the pricing and benchmarks above still apply if you pick one of them. OpenAI's announcements run long, and sometimes you only need the key points.