Overview
GPT-5.6 Sol is the flagship model in OpenAI's GPT-5.6 series, announced on June 26, 2026 alongside two smaller tiers, Terra and Luna. OpenAI describes Sol as its strongest model yet for complex reasoning, advanced coding, agent workflows, and demanding enterprise tasks.
Sol introduces a new max reasoning effort setting that gives the model more time to think, plus an ultra mode that delegates work to subagents to accelerate large, multi-step jobs. The model accepts text and image input and is positioned for agentic coding, scientific research, and high-stakes professional work.
Access at launch is restricted: in coordination with the U.S. government, OpenAI began with a limited preview to around 20 vetted partners through the API and Codex. General availability across ChatGPT, Codex, and the API is planned in the following weeks.
| Released | 2026-06-26 |
|---|---|
| License | Proprietary |
| Weights | API only |
| Modalities | Text, Vision |
| Status | Preview |
Benchmarks
GPT-5.6 Sol vs leading models on Terminal-Bench 2.1, as published by OpenAI at the GPT-5.6 preview launch.
| Benchmark | GPT-5.6 Sol | GPT-5.6 Sol (Ultra) | Claude Mythos 5 | GPT-5.6 Terra | Claude Fable 5 | GPT-5.5 | GPT-5.6 Luna | Claude Opus 4.8 | Gemini 3.1 Pro |
|---|---|---|---|---|---|---|---|---|---|
| Terminal-Bench 2.1 | 88.8% | 91.9% | 88% | 84.3% | 84.3% | 83.4% | 82.5% | 78.9% | 70.7% |
This model's scores
- Terminal-Bench 2.188.8%
Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.
Pricing
| Input | $5.00 / 1M tokens |
|---|---|
| Output | $30.00 / 1M tokens |
Cached input reads receive a 90% discount; cache writes are billed at 1.25x the input rate, with a 30-minute minimum cache life.
Strengths
- Frontier reasoning with a new max reasoning effort tier
- Ultra mode that orchestrates subagents for long-horizon agentic work
- State-of-the-art Terminal-Bench 2.1 score for command-line coding workflows
- Stronger SecureBio biology and cybersecurity reasoning than GPT-5.5
- Improved factuality, with fewer reported hallucinations than the prior flagship
Best for
- Reach for it when an agentic coding task needs the most capable single model and the longest reasoning budget.
- Reach for it for multi-agent workflows where Sol's ultra mode can spin up subagents to parallelise work.
- Reach for it for high-stakes professional reasoning in scientific, biological, or cybersecurity domains.
- Reach for it when factuality matters and the prior GPT-5.5 generation produced too many hallucinated facts.
How to access
| Provider | Model ID |
|---|---|
| OpenAI API ↗ | — |
GPT (Flagship / Thinking) — every version
The full lineage of the GPT (Flagship / Thinking) line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.
| Version | Released | Context | License |
|---|---|---|---|
| GPT-5.6 Sol | 2026-06-26 | — | Proprietary |
| GPT-5.6 Terra | 2026-06-26 | — | Proprietary |
| GPT-5.6 Luna | 2026-06-26 | — | Proprietary |
| GPT-5.5current | 2026-04-23 | 1.05M | Proprietary |
| GPT-5.4 | 2026-03-05 | — | Proprietary |
| GPT-5.2 | 2025-12-11 | — | Proprietary |
| GPT-5.1 | 2025-11-12 | — | Proprietary |
| GPT-5 | 2025-08-07 | — | Proprietary |
| GPT-4o | 2024-05-13 | — | Proprietary |
FAQ
When was GPT-5.6 Sol released?
OpenAI announced GPT-5.6 Sol on June 26, 2026 as a limited preview, alongside the smaller Terra and Luna models in the GPT-5.6 series. Initial access was limited to roughly 20 vetted partner organisations via the API and Codex, with broader rollout planned in the following weeks.
How much does GPT-5.6 Sol cost?
GPT-5.6 Sol is priced at $5.00 per 1M input tokens and $30.00 per 1M output tokens, the same headline rate as GPT-5.5. Cached input reads receive a 90% discount, while cache writes are billed at 1.25x the input rate with a 30-minute minimum cache life.
How does GPT-5.6 Sol compare to Claude Mythos 5 and GPT-5.5?
On Terminal-Bench 2.1, OpenAI reports GPT-5.6 Sol at 88.8% (rising to 91.9% in ultra mode), versus 88.0% for Claude Mythos 5, 84.3% for both GPT-5.6 Terra and Claude Fable 5, and 83.4% for GPT-5.5. OpenAI also reports stronger SecureBio biology and cybersecurity reasoning than GPT-5.5, with reduced hallucination rates.
What is ultra mode in GPT-5.6 Sol?
Ultra mode is a new setting introduced with GPT-5.6 Sol that goes beyond single-agent reasoning by delegating subtasks to subagents to accelerate complex work. Combined with the new max reasoning effort, it lets Sol invest more compute on long-horizon, multi-step problems such as agentic coding.