OpenAI · Released Jul 9, 2026
GPT-5.6 Sol
GPT-5.6 Sol is the flagship model in OpenAI's GPT-5.6 series. It is suited for complex reasoning, coding, and agentic workflows, and is particularly strong at command-line and multi-step coding tasks...
- #1CritPt
- #1OTIS Mock AIME
- Top 3%Chess Puzzles
- Top 7%ARC-AGI-2
- Released
- Jul 9 2026
- Context window
- 1.05M 128K out
- Input / 1M tokens
- $4.00 from $2.00
- Output / 1M tokens
- $20.00 from $10.00
Where to run it
3 providers compared
Cheapest: OpenAI at $2.00 / $10.00 per 1M input / output tokens — 75% less than the priciest option.
| Provider | Input | Output | Cache read | Context | Max output | Uptime | vs. best |
|---|---|---|---|---|---|---|---|
| OpenAI | $2.00 | $10.00 | $0.20 | 1.05M | 128K | 100%24h | Cheapest |
| OpenAI | $4.00 | $20.00 | $0.40 | 1.05M | 128K | 100%24h | +100% |
| Azure | $4.00 | $20.00 | $0.40 | 1.05M | 128K | 99.9%24h | +100% |
| Azureus | $4.40 | $22.00 | $0.44 | 1.05M | 128K | 100%24h | +120% |
| Amazon Bedrockus-east-1 | $4.40 | $22.00 | $0.44 | 1.05M | 128K | 100%24h | +120% |
| Azureeu | $4.40 | $22.00 | $0.44 | 1.05M | 128K | 100%24h | +120% |
| OpenAIfast | $8.00 | $40.00 | $0.80 | 1.05M | 128K | 100%24h | +300% |
Value
Price vs. performance
GPT-5.6 Sol Other OpenAI Other labs Value frontier · GPQA Diamond vs. input price per 1M tokens (log scale) · Drag to pan, pinch or Ctrl + scroll to zoom, double-click to zoom in.
Benchmarks
29 benchmark scores
| Benchmark | Score | Rank | Leader |
|---|---|---|---|
| FrontierSWESoftware engineering · Ultra-long-horizon engineering tasks with a 20-hour budget each | 32.2% | #9of 19 | GPT-6 Astra65.5% |
| DeepSWESoftware engineering · Original long-horizon tasks across active open-source repos | 72.7% | #4of 26 | GPT-6 Astra74.1% |
| CursorBenchSoftware engineering · Ambiguous multi-file tasks drawn from real Cursor sessions | 41.7% | #9of 16 | Claude Opus 5.557.8% |
| FrontierCodeSoftware engineering · Mergeable fixes for hard open-source issues | 47.5% | #12of 41 | Claude Opus 5.554.6% |
| MirrorCodeSoftware engineering · Reimplement whole programs end-to-end without the source | 20.0% | #6of 9 | Claude Opus 5.577.4% |
| GBAEvalSoftware engineering · Build a Game Boy Advance emulator from scratch | 52.6% | #7of 23 | Claude Opus 579.6% |
| GSOSoftware engineering · Optimize real code to make it significantly faster | 76.5% | #4of 20 | Claude Fable 5.188.2% |
| SciCodeSoftware engineering · Research-coding problems from the natural sciences | 57.8% | #14of 93 | Claude Opus 5.566.9% |
| WeirdML v3Software engineering · Unfamiliar ML and data tasks with vague goals | 15.0% | #10of 22 | GPT-6 Astra42.2% |
| APEX-AgentsAgents & computer use · Long professional tasks in banking, consulting and law | 51.4% | #25of 51 | Gemini 4 Argon82.4% |
| OSWorld 2.0Agents & computer use · Long real-world desktop and web tasks | 27.3% | #2of 9 | Claude Opus 531.4% |
| Vending-Bench 2Agents & computer use · Run a simulated vending business for a year (final balance) | $9,619 | #8of 61 | GPT-6 Astra$15,515 |
| PostTrainBenchAgents & computer use · Post-train small models under a fixed compute budget | 36.2% | #2of 11 | Claude Fable 541.8% |
| GDP.pdfWriting & content · Answer questions about real professional PDFs | 30.7% | #4of 38 | GPT-6.1 Sol35.0% |
| SimpleQA VerifiedWriting & content · Factual accuracy on 1,000 short questions | 69.7% | #8of 66 | Gemini 4 Argon76.0% |
| WebDev ArenaDesign & visual · People vote on which model builds the better website (Elo) | 1617 | #16of 88 | Claude Opus 5.51820 |
| Blueprint-Bench 2Design & visual · Turn apartment photos into 2D floor plans | 33.6% | #10of 32 | Gemini 4 Argon54.4% |
| Furniture AssemblyDesign & visual · Spot mistakes in photos of IKEA builds | 56.7% | #9of 34 | Claude Opus 5.583.3% |
| GPQA DiamondReasoning & knowledge · PhD-level biology, chemistry and physics questions | 93.5% | #13of 92 | Gemini 4 Argon96.4% |
| ARC-AGI-2Reasoning & knowledge · Few-shot abstract reasoning on grid puzzles | 92.5% | #4of 63 | GPT-6 Astra95.0% |
| ARC-AGI-1Reasoning & knowledge · Pattern induction from input–output examples | 97.5% | #8of 63 | Gemini 4 Argon98.5% |
| SimpleBenchReasoning & knowledge · Common-sense trick questions about space, time and people | 71.7% | #9of 43 | Claude Fable 581.9% |
| CritPtReasoning & knowledge · Open-ended, research-level physics problems | 32.3% | #1of 96 | This model |
| Chess PuzzlesReasoning & knowledge · Find the single best move in 100 new puzzles | 68.0% | #2of 70 | GPT-6 Astra76.0% |
| LMCAReasoning & knowledge · Judge the quality of conceptual arguments like experts | 59.2% | #9of 96 | Claude Sonnet 5.568.3% |
| FrontierMathMaths · Expert-written problems, undergrad to early research (tiers 1–3) | 89.1% | #6of 72 | GPT-6.1 Sol93.7% |
| FrontierMath Tier 4Maths · Exceptionally hard research-level problems | 82.9% | #7of 62 | GPT-6.1 Sol100.0% |
| ProofBenchMaths · Graduate maths proofs verified in Lean 4 | 83.0% | #11of 70 | Claude Sonnet 5.5100.0% |
| OTIS Mock AIMEMaths · Competition-style problems harder than MATH Level 5 | 100.0% | #1of 90 | Gemini 4 Argon100.0% |
Specs
Details
- Model ID
openai/gpt-5.6-sol- Developer
- OpenAI
- Release date
- Jul 9, 2026 ·
- Context window
- 1,050,000 tokens
- Max output
- 128,000 tokens
- API price
- $4.00 / $20.00 per 1M input / output tokens
- Providers
- 3
- Input types
- file, image, text
- Reasoning
- Yes
- Tool calling
- Yes
- Open weights
- No
- Free tier
- No
Release date, prices and providers from OpenRouter's catalog; buying directly from a provider can cost a little differently. Uptime is the share of OpenRouter requests each provider served over the last day. Benchmark scores are the best result per benchmark from Epoch AI, licensed CC BY 4.0. Rank counts every model tracked here with a score.