ChatGBot
ModelsFeaturesApp BuilderPricingFAQ
Sign inStart free
← Back to blogNovember 18, 2026

GPT-6.2 Astra: OpenAI's Flagship Gets Cheaper and Smarter

By the ChatGBot Team · Published November 18, 2026

OpenAI released GPT-6.2 Astra on November 17, 2026, together with GPT-6.2 Sol. Astra is OpenAI's flagship for the hardest work: debugging large codebases, scientific analysis and agents that run for a long time. It is also 20% cheaper than GPT-6 Astra. Here is what changed, how it scores, what it costs and where the critics disagree.

ClaudeGPTDeepSeekQwenGLM

Every top AI in one app

GPT, Claude, DeepSeek, Qwen and GLM. One subscription.

Try ChatGBot →

GPT-6.2 Astra at a glance

GPT-6.2 Astra
Made byOpenAI
ReleasedNovember 17, 2026
API model IDgpt-6.2-astra
ReplacesGPT-6 Astra
Context window1,050,000 tokens
Max output128,000 tokens
InputText and images
OutputText
Knowledge cutoffMay 2026
API price per 1M tokens$8 input / $40 output
In ChatGPTChatGPT Work and Codex (Plus, Pro, Enterprise, Edu)

What is GPT-6.2 Astra?

OpenAI calls Astra its model for your hardest projects, from start to finish. It is built to stay on a big task for a long time: planning, writing code, running tools and checking its own work. OpenAI also calls it the best model in the world for scientists. Its everyday sibling is GPT-6.2 Sol, which is much cheaper and almost as strong on many tasks.

Note that the new models arrive in ChatGPT Work and Codex. OpenAI says the rollout does not add them to regular ChatGPT chat.

What is new

  • Lower price. Input and output are 20% cheaper than GPT-6 Astra, and cached input is 60% cheaper ($0.40 instead of $1).
  • Fewer tokens. OpenAI says Astra uses 22% fewer output tokens on GDPval v2 while scoring 6.2 points higher.
  • Better at science and agents. Big jumps on Terminal-Bench-Science and LifeVal (see the table below).
  • Tools. In the API it supports web search, file search, code interpreter, computer use, MCP and a multi-agent beta.

Benchmarks

Test (OpenAI, max reasoning)GPT-6 AstraGPT-6.2 Astra
Agents' Last Exam v1.173.1%77.6%
Terminal-Bench-Science 0.168.1%77.1%
FrontierCode 2.042.8%44.2%
LifeVal79.8%86.1%

Outside tests: Astra scores 62 on the Artificial Analysis Intelligence Index, second behind Claude Fable 5.2 at 65. The ARC Prize reports 95.8% on ARC-AGI-2. OpenAI also says Astra beats Fable 5.2 by 5.5 points on Agents' Last Exam at about 5% of the cost.

What critics say

Not every test is glowing. Artificial Analysis measured a 59% hallucination rate on hard factual questions at max effort. Epoch AI found that Astra still trails Claude Fable 5.2 on real-world research tasks and barely improves on GPT-6 Astra there. Ethan Mollick called it uneven and harder to steer, though able to finish an ambitious project. Most of the headline gains are OpenAI's own numbers.

Safety

OpenAI's system card rates both new models at its Critical level for cybersecurity and High for biology and chemistry. Astra is the first model OpenAI has released that is rated High for AI self-improvement. As a result, OpenAI has tightened default limits on risky security help and added extra checks for wider access.

Price and how to get access

Per 1M tokensInputCached inputOutput
GPT-6.2 Astra$8$0.40$40
GPT-6 Astra$10$1.00$50
GPT-6.2 Astra, long context (over 272K)$16$0.80$60
GPT-6.2 Astra, Batch$4$0.20$20
  • ChatGPT: ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu. Enterprise and Edu admins must enable it first.
  • API: model ID gpt-6.2-astra.
  • GPT-6 Astra is still listed in the API at its old price.

Should you use it?

If you run long coding projects, research or agents and want OpenAI's strongest model, Astra is now cheaper to try. For most everyday writing, coding and questions, GPT-6.2 Sol gives you much of the quality at a fifth of the price. And for some research tasks, Claude Fable 5.2 is still ahead. Comparing models on your own task is the safest way to choose. For more launches this month, see our November 2026 AI model releases roundup.

Use top models in one place

ClaudeGPTDeepSeekQwenGLM

Ready to try it yourself?

Switch between top AI models in one chat.

Try ChatGBot →

Frequently asked questions

What is GPT-6.2 Astra?

GPT-6.2 Astra is OpenAI's most capable model for hard, long projects such as large codebases, scientific analysis and long-running agents. It was released on November 17, 2026.

How much does GPT-6.2 Astra cost?

In the API it costs $8 per million input tokens and $40 per million output tokens, 20% less than GPT-6 Astra. Cached input costs $0.40 per million tokens.

Who can use GPT-6.2 Astra?

It is available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API as gpt-6.2-astra. Enterprise and Edu admins need to turn it on first.

Is GPT-6.2 Astra better than Claude Fable 5.2?

OpenAI says Astra scores 5.5 points higher on Agents' Last Exam at about 5% of the cost. But Artificial Analysis still ranks Claude Fable 5.2 higher on its Intelligence Index, 65 to 62.

What is the GPT-6.2 Astra context window?

1,050,000 tokens, with up to 128,000 output tokens. It reads text and images and answers in text.

Can I use GPT models in ChatGBot?

Yes. ChatGBot offers GPT, Claude, DeepSeek, Qwen and GLM models in one app, so you can compare answers side by side.

One subscription. Multiple leading AI models.

Start free

Blog

  • Gemini 4 Neon: Google's New Mid-Tier Model, Explained
  • GPT-6.2 Sol: OpenAI's Everyday Model Gets Better at the Same Price
  • Qwen4-Max: Alibaba's New Flagship Model, Explained
  • Gemini 4 Hydrogen: Google's Fast New Model, Explained
ChatGBot

One subscription for every leading AI model: chat, images, web search, documents, Canvas and the App Builder.

support@chatgbot.ai

Product

ModelsFeaturesApp BuilderPricingFAQSign in

Legal

Terms of UsePrivacy PolicyRefund Policy
BlogGemini 4 Neon: Google's New Mid-Tier Model, ExplainedGPT-6.2 Sol: OpenAI's Everyday Model Gets Better at the Same PriceGPT-6.2 Astra: OpenAI's Flagship Gets Cheaper and SmarterQwen4-Max: Alibaba's New Flagship Model, ExplainedAll articles

© 2026 ChatGBot. Operated by 42 Dijital Yazılım Limited Şirketi.

Chatgbot is an independent AI interface. It is not affiliated with, endorsed by, or sponsored by OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen, GLM, or other model providers. Model names and trademarks belong to their respective owners. Model availability may vary by plan, region, and provider access.