Questa sezione non è ancora tradotta. Il contenuto seguente è mostrato in inglese, come pubblicato. Leggi la versione inglese.
OpenAI · Text model · GA
OpenAI: for complex reasoning, coding, computer use, research and document creation developers.openai.com (opens in a new tab)
Compare with similar models Open the Model Finder
Figures checked 1 Nov 2026. A dash means not published on the sources we checked, so it stays blank rather than estimated. AA index and speed are from Artificial Analysis at the max setting.
OpenAI names these settings reasoning.effort. Price per token is the same at every level. Higher effort uses more reasoning tokens, billed as output. developers.openai.com (opens in a new tab)
| Level | When to use it, per OpenAI | AA index | Tokens/s |
|---|---|---|---|
| low | Efficient reasoning with a modest latency increase. Data analysis, drafting, execution-oriented coding and support chat developers.openai.com (opens in a new tab) | 43 artificialanalysis.ai (opens in a new tab) | 58 artificialanalysis.ai (opens in a new tab) |
| medium | When quality and reliability matter. The default for most workloads and a balanced point for latency, performance and cost developers.openai.com (opens in a new tab) | 51 artificialanalysis.ai (opens in a new tab) | 57 artificialanalysis.ai (opens in a new tab) |
| high | Hard reasoning, complex debugging, deep planning and high-value tasks where quality matters more than latency developers.openai.com (opens in a new tab) | 55 artificialanalysis.ai (opens in a new tab) | 63 artificialanalysis.ai (opens in a new tab) |
| xhigh | Deep research, asynchronous and long agentic runs. Only when your evals show a clear benefit for the extra latency and cost developers.openai.com (opens in a new tab) | 55 artificialanalysis.ai (opens in a new tab) | 66 artificialanalysis.ai (opens in a new tab) |
| max | Maximum reasoning for your most complex tasks. If you use xhigh, test whether max does better developers.openai.com (opens in a new tab) | 57 artificialanalysis.ai (opens in a new tab) | 78 artificialanalysis.ai (opens in a new tab) |
Our rule of thumb: Simple or speed-critical work: low. Standard work: the lab's default. Complex work: high. Use xhigh or max only when your own tests show a gain, as OpenAI and Anthropic both advise. developers.openai.com (opens in a new tab)platform.claude.com (opens in a new tab)
Artificial Analysis Intelligence Index at each setting it tested. Levels it has not tested are left out.
| Level | AA index | Tokens/s |
|---|---|---|
| low | 43 | 58 |
| medium | 51 | 57 |
| high | 55 | 63 |
| xhigh | 55 | 66 |
| max | 57 | 78 |
Output tokens a second measured by Artificial Analysis at each setting.
| Level | Tokens/s |
|---|---|
| low | 58 |
| medium | 57 |
| high | 63 |
| xhigh | 66 |
| max | 78 |
Highlighted point is GPT-6 Astra. Grey points are the other 26 text models with both figures. Index at each model's best tested setting.
| Model | Index | Blended US$/1M |
|---|---|---|
| GPT-6 Astra | 57 | 20.000 |
| GPT-6.1 Sol | 55 | 4.000 |
| GPT-6.1 Terra | 51 | 2.000 |
| GPT-6 Luna | 28 | 0.200 |
| Claude Fable 5.1 | 50 | 20.000 |
| Claude Opus 5.5 | 58 | 8.000 |
| Claude Sonnet 5.5 | 54 | 4.000 |
| Claude Haiku 5.5 | 36 | 0.200 |
| Gemini 4 Carbon | 59 | 4.000 |
| Gemini 4 Argon | 50 | 4.000 |
| Gemini 3.8 Flash | 31 | 1.500 |
| Gemini 3.5 Flash-Lite | 12 | 0.850 |
| Grok 4.7 | 35 | 3.000 |
| Muse Wonder | 43 | 8.000 |
| Mistral Large 4 | 26 | 1.050 |
| Mistral Small 4 | 4 | 0.262 |
| DeepSeek V4.1 Flash 1017 | 32 | 0.525 |
| DeepSeek V4 Pro | 23 | 1.980 |
| Qwen3.8-Max | 34 | 3.000 |
| Kimi K3.1 Pro | 46 | 4.000 |
| GLM-5.3 | 32 | 2.150 |
| GLM-5.3-Flash | 28 | 0.237 |
| MiniMax M3.1 Flash | 29 | 0.262 |
| MiniMax M3 | 19 | 0.525 |
| MiMo-V2.6-Pro | 30 | 0.544 |
| MiMo-V2.6-Flash | 27 | 0.175 |
| Step 5 Preview | 29 | 1.425 |
US$ per 1M tokens, standard tier, short prompts. Output left blank where not published.
| Model | Input US$/1M | Output US$/1M |
|---|---|---|
| GPT-6 Astra | 10 | 50 |
| Claude Fable 5.1 | 10 | 50 |
| Claude Opus 5.5 | 4 | 20 |
| Gemini 4 Carbon | 2 | 10 |
| GPT-6.1 Sol | 2 | 10 |
| Lab | OpenAI |
|---|---|
| Type | Text |
| Status | GA developers.openai.com (opens in a new tab) |
| Released | 4 Sep 2026 models.dev (opens in a new tab) |
| API access | Yes developers.openai.com (opens in a new tab) |
| API model ID | gpt-6-astra developers.openai.com (opens in a new tab) |
| API rate limits | Build: 5,000 requests a minute, 1,000,000 tokens a minute. Launch: 10,000 requests a minute, 4,000,000 tokens a minute. Grow: 15,000 requests a minute, 40,000,000 tokens a minute developers.openai.com (opens in a new tab)Standard tier limits. Your tier rises with spend |
| Consumer app plans | Pro chatgpt.com (opens in a new tab)ChatGPT Pro: "Pro reasoning powered by GPT-6 Astra" |
| Price at each level | Same per-token price at every level. developers.openai.com (opens in a new tab) |
| Input per 1M tokens | US$10 developers.openai.com (opens in a new tab)Up to 272K input. US$20 above |
| Output per 1M tokens | US$50 developers.openai.com (opens in a new tab)Up to 272K input. US$75 above |
| Cached input | US$1 developers.openai.com (opens in a new tab) |
| Batch in / out | US$5 / US$25 developers.openai.com (opens in a new tab) |
| Other pricing | – |
| Context window | 1.05M tokens developers.openai.com (opens in a new tab) |
| Max output | 128K tokens developers.openai.com (opens in a new tab) |
| Input types | text, image developers.openai.com (opens in a new tab) |
| Output types | text |
| Weights | Closed developers.openai.com (opens in a new tab) |
| Reasoning | Yes developers.openai.com (opens in a new tab) |
| Tool use | Yes developers.openai.com (opens in a new tab) |
| Artificial Analysis index | 57 (max) artificialanalysis.ai (opens in a new tab) |
| Output speed | 78 tokens/s artificialanalysis.ai (opens in a new tab) |
| Time to first token | 270.03 s artificialanalysis.ai (opens in a new tab) |
| Reasoning levels | low, medium, high, xhigh, max developers.openai.com (opens in a new tab) |
| Lab benchmark claims | – |
| Limitations |
|
OpenAI: Remote MCP servers as a tool in the Responses API developers.openai.com (opens in a new tab). These official remote MCP servers fit that rule. We have not tested each server with this model. A server that publishes a client list is left out when this lab is not on it.
Same type, closest on quality index and blended price. A similarity, not a ranking.
Compare GPT-6 Astra with Claude Fable 5.1 and Claude Opus 5.5
All 109 AI models · AI agents · How we check sources
OpenAI 
CNBC confirmed OpenAI is now seeking up to US$20bn, down from about US$30bn, and the FT put the post-money valuation at US$1.25tn. The round is still open. Less fresh equity means more scrutiny of every gigawatt OpenAI has signed for.
Source: CNBC (opens in a new tab)
OpenAI
OpenAI says Terra matches GPT-6 Astra on Agents' Last Exam at a tenth of the price per token and finishes tasks over 3x faster.
Source: OpenAI (opens in a new tab)
