GPT-6.2 Astra is OpenAI's flagship model in the GPT-6.2 series, designed for the most demanding projects. It is a proprietary model available through OpenAI's API. Prices listed here are from OpenAI's official pricing page (developers.openai.com/api/docs/pricing, checked Nov 18, 2026): $8.00 per million input tokens and $40.00 per million output tokens at the standard tier, with a cached input rate of $0.40/1M. A long-context tier is also available at $16.00/$60.00 input/output with a $0.80/1M cached rate.
GPT-6.2 Astra supports a 1,050,000-token context window and a maximum output of 128,000 tokens. It supports vision input, function calling, and streaming. The model is not a reasoning model in the chain-of-thought sense; for higher-quality reasoning on the same underlying model, see GPT-6.2 Astra Pro, which is served with reasoning.mode set to pro and priced identically by OpenAI.
The GPT-6.2 family sits above GPT-6.1 in OpenAI's lineup. Astra is the flagship variant; Sol and Sol Pro are the more cost-efficient variants at $2.00/$10.00 per million tokens. All four GPT-6.2 variants share the same 1,050,000-token context window and 128,000-token output limit.
Data source for prices: OpenAI official pricing page, checked Nov 18, 2026. Data source for listing: OpenRouter listing. Data confidence: Partial — benchmark scores are not available for this model at this time.
GPT-6.2 Astra is suited to applications requiring the highest capability in the GPT-6.2 family: complex multi-step reasoning, long-document analysis, vision tasks, and agentic workflows. Teams with cost constraints may find GPT-6.2 Sol or Sol Pro a better fit at one-quarter the price.
Input price vs 143 frontier models tracked. Avg: $5.58/1M.
| Model ID | gpt-6.2-astra |
| Provider | OpenAI |
| Model Family | GPT-6.2 |
| Tier | Frontier |
| Context Window | 1,050,000 tokens (1.1M) |
| Max Output Tokens | 128,000 tokens |
| Input Price | $8 / 1M tokens |
| Output Price | $40 / 1M tokens |
| Cached Input Price | $0.4 / 1M tokens |
| Throughput | — |
| Latency (P50) | — |
| Intelligence Score | — |
| LiveBench Score | — |
| Cost per Task (LiveBench) | — |
| Data Confidence | Partial |
| HLE Score | — |
| GPQA Diamond | — |
| SWE-bench Score | — |
| OSWorld Score | — |
| BrowseComp Score | — |
| Terminal-Bench Score | — |
| AutoBench Score | — |
| Parameters | — |
| Vision Support | Yes |
| Function Calling | Yes |
| Streaming | Yes |
| Reasoning Model | No |
| Open Weights | No |
| Data Source | Static (verified) |
| Last Updated | 2026-11-18 |
Self-hosting gpt-6.2-astra? These GPUs are well-matched for this model's compute and memory requirements.