GPT-5.6 Sol
OpenAIJul 9, 2026#2
Flagship of OpenAI's three-tier GPT-5.6 family (Sol · Terra · Luna), aimed at complex reasoning and professional workflows. Sets public SOTA on BrowseComp and OSWorld 2.0.
- Status
- Generally available
- License
- Proprietary API
- Context window
- 1.1M
- Max output
- 128K
- Input · $/M
- $5
- Output · $/M
- $30
- Cached input · $/M
- $0.5
- Blended · $/M
- $11.25
- Parameters
- Undisclosed
- Inputs
- Text + images, audio
Benchmarks
Composite 78.3· mean of published scores
Reasoning & knowledge
GPQA DiamondLeader
93.6%
MMLU-Pro
no public score
Humanity's Last Exam
no public score
ARC-AGI-2
no public score
Coding
SWE-bench Verified
no public score
SWE-bench Pro
64.6%
dataset best 80LiveCodeBench
no public score
Agents & computer use
Terminal-Bench 2
no public score
OSWorld 2.0Leader
62.6%
BrowseCompLeader
92.2%
Math
AIME 2026
no public score
Known strengths
- Verified GPQA leader
- SOTA agentic browsing
- Top computer-use score
Caveats
- OpenAI reports SWE-bench Pro rather than SWE-bench Verified
- Vendor-reported scores