How much does GPT-5.6 Sol cost?

GPT-5.6 Sol is priced at $5 per million input tokens and $30 per million output tokens on the OpenAI API. Terra sits at $2.50 / $15 per million and Luna at $1 / $6 per million. The pricing lands Terra roughly half of GPT-5.5 for similar performance and Luna as OpenAI's cheapest text model.

How does GPT-5.6 Sol compare to Claude Mythos 5?

GPT-5.6 Sol scores 88.8% on Terminal-Bench 2.1 versus Claude Mythos 5 at 88.0%, and Sol Ultra pushes to 91.9% by spawning parallel subagents. On price, Sol matches Mythos 5's input rate at $5 per million tokens but comes in cheaper on output at $30 versus Anthropic's $75.

What is Sol Ultra mode?

Sol Ultra is a new max-reasoning setting inside GPT-5.6 Sol that decomposes a request and spawns parallel subagent processes trained to coordinate mid-task. Sol Ultra spends more compute per prompt in exchange for a Terminal-Bench 2.1 lift from 88.8% to 91.9% — about three points — on long-horizon coding tasks.

Why was GPT-5.6 delayed until July 9?

GPT-5.6 was previewed to trusted partners on June 26 while the Trump administration ran an added review of the model. The White House cleared the wider rollout after more testing and meetings with OpenAI, so Sol, Terra, and Luna went public on July 9 across ChatGPT, the API, and Codex.

What is ChatGPT Work?

ChatGPT Work is a new workplace agent OpenAI shipped alongside GPT-5.6, powered by the Sol tier. ChatGPT Work connects to Slack, Teams, Google Drive, SharePoint, email, calendars, and CRMs, and can run for hours on scheduled tasks that produce finished documents, spreadsheets, presentations, and web apps.

OpenAI · 2026-07-09 · major

GPT-5.6 goes public — Sol, Terra, and Luna clear the White House gate

OpenAI released GPT-5.6 to everyone on July 9 after federal reviewers cleared it. Sol, Terra, and Luna are live in ChatGPT, the API, and Codex, and a new Sol Ultra mode runs subagents in parallel to hit 91.9% on Terminal-Bench 2.1.

OpenAI branding used in Engadget's coverage of the GPT-5.6 public rollout on July 9, 2026 — Engadget

OpenAI's frontier model line goes broad after the government finishes its second look.

Key specs

Terminal bench 2.1 (sol ultra)	91.9%
Sol input	$5 / 1M tokens
Sol output	$30 / 1M tokens

Quick facts

Maker	OpenAI
Tiers	Sol, Terra, Luna (plus Sol Ultra)
Availability	ChatGPT, API, Codex — public July 9
Preview → GA gap	13 days (previewed June 26)
What's new	Sol Ultra parallel subagents; ChatGPT Work agent; Sites beta
Also launched today	ChatGPT Work, ChatGPT desktop app, Sites beta

Benchmarks

Terminal-Bench 2.1

GPT-5.6 Sol Ultra		91.9%
GPT-5.6 Sol		88.8%
Claude Mythos 5		88%
Claude Fable 5		84.3%
Claude Opus 4.8		78.9%

source ↗

Pricing

Sol input	$5.00 / 1M tokens
Sol output	$30.00 / 1M tokens
Terra input	$2.50 / 1M tokens
Terra output	$15.00 / 1M tokens
Luna input	$1.00 / 1M tokens
Luna output	$6.00 / 1M tokens

source ↗

What is it?

GPT-5.6 is OpenAI's July 9 general-availability release of Sol, Terra, and Luna, the three-tier successor to GPT-5.5. Sol is the frontier reasoning tier, Terra a balanced everyday tier at roughly half GPT-5.5's price, and Luna the cheapest option at $1 in / $6 out per million tokens. All three now run inside ChatGPT, the OpenAI API, and Codex, ending the trusted-partners-only preview that started June 26.

How does it work?

The Sol tier ships with a new Sol Ultra mode that decomposes a prompt and spawns parallel subagent processes trained to coordinate mid-task, rather than each subagent working in isolation. Sol Ultra spends more compute per prompt but lifts Terminal-Bench 2.1 from Sol's 88.8% to 91.9%. Terra keeps GPT-5.5-class performance at roughly half the price, and Luna is optimized for high-throughput classification, routing, and small edits.

Why does it matter?

GPT-5.6 marks the first frontier launch that OpenAI held back for a federal-government review, so July 9 is the day the model actually reaches the public. Sol Ultra's 91.9% Terminal-Bench 2.1 clears Anthropic's Claude Mythos 5 at 88.0% on long-horizon coding tasks, and Terra's price cut aims squarely at the middle of the market that GPT-5.5 owned. ChatGPT Work, a new workplace agent that runs on Sol, shipped the same day.

Who is it for?

OpenAI API customers, ChatGPT teams, and Codex users who were blocked from Sol during the June 26 trusted-partner preview.

Frequently asked questions

How much does GPT-5.6 Sol cost?: GPT-5.6 Sol is priced at $5 per million input tokens and $30 per million output tokens on the OpenAI API. Terra sits at $2.50 / $15 per million and Luna at $1 / $6 per million. The pricing lands Terra roughly half of GPT-5.5 for similar performance and Luna as OpenAI's cheapest text model.
How does GPT-5.6 Sol compare to Claude Mythos 5?: GPT-5.6 Sol scores 88.8% on Terminal-Bench 2.1 versus Claude Mythos 5 at 88.0%, and Sol Ultra pushes to 91.9% by spawning parallel subagents. On price, Sol matches Mythos 5's input rate at $5 per million tokens but comes in cheaper on output at $30 versus Anthropic's $75.
What is Sol Ultra mode?: Sol Ultra is a new max-reasoning setting inside GPT-5.6 Sol that decomposes a request and spawns parallel subagent processes trained to coordinate mid-task. Sol Ultra spends more compute per prompt in exchange for a Terminal-Bench 2.1 lift from 88.8% to 91.9% — about three points — on long-horizon coding tasks.
Why was GPT-5.6 delayed until July 9?: GPT-5.6 was previewed to trusted partners on June 26 while the Trump administration ran an added review of the model. The White House cleared the wider rollout after more testing and meetings with OpenAI, so Sol, Terra, and Luna went public on July 9 across ChatGPT, the API, and Codex.
What is ChatGPT Work?: ChatGPT Work is a new workplace agent OpenAI shipped alongside GPT-5.6, powered by the Sol tier. ChatGPT Work connects to Slack, Teams, Google Drive, SharePoint, email, calendars, and CRMs, and can run for hours on scheduled tasks that produce finished documents, spreadsheets, presentations, and web apps.

Try it

gpt-5-6-sol / gpt-5-6-terra / gpt-5-6-luna on the OpenAI API.