What is GPT-5.6 Sol?
As of 2026-08-12, api model id: gpt-5.6-sol.
Model reference
GPT-5.6 Sol
This reference covers the API-only flagship capability tier in a model family whose tiers advance on independent cadences. It keeps sibling pricing changes and hosted availability separate from capability claims.
GPT-5.6 Sol is the vendor reference for gpt-5.6-sol, as of 2026-08-12. Context window: 1,050,000, as of 2026-08-12. RunInfra has not measured this model; every figure below belongs to its named source, cited and dated.
Cited specifications
| Group | Fact | Source-cited display value | Source and date |
|---|---|---|---|
| Identity | API model id | gpt-5.6-sol | SourceRetrieved 2026-08-12 |
| Identity | Series launch | We're launching the GPT-5.6 family of models for general availability following our limited preview: our new flagship, Sol, alongside Terra, a balanced model for everyday work, and Luna, our most cost-efficient model. | SourceRetrieved 2026-08-12 |
| Identity | Series launch publication date | July 9, 2026 | SourceRetrieved 2026-08-12 |
| Identity | Series naming system | In this new naming system introduced with GPT-5.6, the number identifies a model's generation, while Sol, Terra, and Luna identify durable capability tiers that can advance on their own cadence. | SourceRetrieved 2026-08-12 |
| Identity | Preview publication date | June 26, 2026 | SourceRetrieved 2026-08-12 |
| Identity | Vendor positioning | GPT-5.6 Sol sets a new standard for both intelligence and efficiency, achieving state-of-the-art results across coding, knowledge work, cybersecurity, and science while outperforming previous and competing frontier models with fewer tokens and at lower estimated cost. | SourceRetrieved 2026-08-12 |
| Architecture | - | - | - |
| Context | Context window | 1,050,000 | SourceRetrieved 2026-08-12 |
| Context | Maximum input | 922,000 tokens | SourceRetrieved 2026-08-12 |
| Context | Maximum output | 128,000 tokens | SourceRetrieved 2026-08-12 |
| Context | Knowledge cutoff | February 16, 2026 | SourceRetrieved 2026-08-12 |
| Modalities | Input and output | text and image input; text output | SourceRetrieved 2026-08-12 |
| Modalities | Reasoning effort | none, low, medium (default), high, xhigh, and max | SourceRetrieved 2026-08-12 |
| Modalities | Maximum reasoning effort | max gives GPT-5.6 even more time than xhigh to reason and explore alternatives, run checks, and revise its approach. | SourceRetrieved 2026-08-12 |
| Modalities | Ultra mode | ultra goes further by coordinating four agents in parallel by default, trading higher token use for stronger results and faster time-to-result on demanding tasks. | SourceRetrieved 2026-08-12 |
| License | - | - | - |
| Pricing | OpenAI standard input | $5.00 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | OpenAI standard cached input | $0.50 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | OpenAI standard cache write | $6.25 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | OpenAI standard output | $30.00 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | OpenAI long-context input | $10.00 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | OpenAI long-context cached input | $1.00 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | OpenAI long-context cache write | $12.50 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | OpenAI long-context output | $45.00 per 1M tokens | SourceRetrieved 2026-08-12 |
| Pricing | Long-context threshold | Prompts with >272K input tokens are priced at 2x input and 1.5x output | SourceRetrieved 2026-08-12 |
| Pricing | Cache economics | cache writes are billed at 1.25x the model's uncached input rate, while cache reads continue to receive the 90% cached-input discount. | SourceRetrieved 2026-08-12 |
| Pricing | Sibling price cut | Update on July 30, 2026: OpenAI reduced the price of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%. | SourceRetrieved 2026-08-12 |
| Pricing | OpenRouter listing | $5.00/M input tokens and $30.00/M output tokens; listed July 9, 2026 | SourceRetrieved 2026-08-12 |
| Availability | Launch availability | GPT-5.6 is available starting today across ChatGPT, Codex, and the OpenAI API. | SourceRetrieved 2026-08-12 |
| Availability | Tier rate limits | Tier 5: 15,000 RPM and 40,000,000 TPM | SourceRetrieved 2026-08-12 |
| Availability | Sol Pro | Pro and Enterprise users can also select GPT-5.6 Sol Pro for the highest-quality results on complex tasks. | SourceRetrieved 2026-08-12 |
What the vendor says is new
"We're launching the GPT-5.6 family of models for general availability following our limited preview: our new flagship, Sol, alongside Terra, a balanced model for everyday work, and Luna, our most cost-efficient model."
SourceRetrieved 2026-08-12
"ultra goes further by coordinating four agents in parallel by default, trading higher token use for stronger results and faster time-to-result on demanding tasks."
SourceRetrieved 2026-08-12
Vendor-claimed benchmarks
These results are vendor-claimed, not independently measured by RunInfra.
| Benchmark | Vendor-claimed value | Source and date |
|---|---|---|
| Agents' Last Exam | 52.7% | SourceRetrieved 2026-08-12 |
| Artificial Analysis Intelligence Index v4.1 | 58.9 | SourceRetrieved 2026-08-12 |
| Artificial Analysis Coding Agent Index v1.1 | 80 | SourceRetrieved 2026-08-12 |
| SWE-Bench Pro | 64.6% | SourceRetrieved 2026-08-12 |
| DeepSWE v1.1 | 72.7% | SourceRetrieved 2026-08-12 |
| Terminal-Bench 2.1 | 88.8% | SourceRetrieved 2026-08-12 |
| BrowseComp | 90.4% | SourceRetrieved 2026-08-12 |
| OSWorld 2.0 | 62.6% | SourceRetrieved 2026-08-12 |
| GPQA Diamond | 94.6% | SourceRetrieved 2026-08-12 |
| Capture-the-Flag | 96.7% | SourceRetrieved 2026-08-12 |
| FrontierMath Tiers 1-3 | 89% | SourceRetrieved 2026-08-12 |
| MMMU Pro, no tools | 83% | SourceRetrieved 2026-08-12 |
Serving support
Listed rows have a cited upstream support signal. They are not RunInfra measurements.
No open-engine serving signal applies to this model. It is served only through the vendor's own API.
Serving concepts
What RunInfra measures when we measure it
RunInfra has not measured this model yet. When measurement is published, the record will state throughput, latency, memory, quality, serving conditions, and reproducible evidence.
Questions about this model reference
What does the GPT-5.6 Sol API cost?
As of 2026-08-12, openai standard input: $5.00 per 1M tokens. As of 2026-08-12, openai standard cached input: $0.50 per 1M tokens. As of 2026-08-12, openai standard cache write: $6.25 per 1M tokens. As of 2026-08-12, openai standard output: $30.00 per 1M tokens. As of 2026-08-12, openai long-context input: $10.00 per 1M tokens. As of 2026-08-12, openai long-context cached input: $1.00 per 1M tokens. As of 2026-08-12, openai long-context cache write: $12.50 per 1M tokens. As of 2026-08-12, openai long-context output: $45.00 per 1M tokens. As of 2026-08-12, long-context threshold: Prompts with >272K input tokens are priced at 2x input and 1.5x output. As of 2026-08-12, cache economics: cache writes are billed at 1.25x the model's uncached input rate, while cache reads continue to receive the 90% cached-input discount. As of 2026-08-12, sibling price cut: Update on July 30, 2026: OpenAI reduced the price of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%. As of 2026-08-12, openrouter listing: $5.00/M input tokens and $30.00/M output tokens; listed July 9, 2026.
Can I run GPT-5.6 Sol myself?
As of 2026-08-12, launch availability: GPT-5.6 is available starting today across ChatGPT, Codex, and the OpenAI API.
RunInfraby RightNow© 2026 RunInfra. All rights reserved.
Ask AI about RunInfra
