# GPT-6.2 Astra and Sol — OpenAI's new flagship pair, Astra 20% cheaper

> OpenAI's first Astra upgrade since September does more with fewer tokens and costs less per token.

- Org: OpenAI
- Date: 2026-11-17
- Category: model
- Source: https://9to5mac.com/2026/11/17/openai-upgrades-chatgpt-and-codex-with-new-gpt-6-2-astra-and-sol-models/
- Canonical page: https://ai-tldr.dev/releases/openai-gpt-6-2-astra-sol/

## In short

GPT-6.2 Astra and GPT-6.2 Sol are OpenAI's new frontier and everyday models. Both use fewer output tokens per task. Astra drops to $8 / $40 per 1M tokens, 20% below GPT-6 Astra; Sol stays at $2 / $10.

## What it is

GPT-6.2 Astra is OpenAI's new frontier model for the hardest projects, and GPT-6.2 Sol is its efficient model for everyday work. OpenAI says both give better results on long, demanding tasks while writing fewer output tokens, so people get more done before they hit usage limits. Astra is the first new Astra since GPT-6 Astra; the planned GPT-6.1 Astra was cancelled in September.

## How it works

Gains in visual reasoning and computer use let the GPT-6.2 models follow slide templates, build spreadsheets that reconcile, and work in CAD and simulation tools; OpenAI shows Astra building a car-chassis simulation and an RC-buggy parts catalog. In Codex, OpenAI says Astra writes cleaner code and more focused tests in existing projects. Both models take text and images, have a 1.05M-token context, and run on the Chat Completions, Responses and Batch APIs.

## Why it matters

Per OpenAI, Astra beats GPT-6 Astra by 5.2 points on Agents' Last Exam at under a third of the cost, and Sol comes within 0.5 points of Astra on Cognition's FrontierCode 2.0 at about a fifth of the cost. Lower token prices plus fewer tokens per task cut the bill for long agent runs. OpenAI also rates both models Critical for cybersecurity, so they ship with tighter cyber safeguards that block some legitimate security research.

## Who it's for

developers and teams running long coding and office agents

## Try it

Call model gpt-6.2-astra or gpt-6.2-sol in the OpenAI API, or pick them in ChatGPT Work and Codex

## Quick facts

- Maker: OpenAI
- Model IDs: gpt-6.2-astra, gpt-6.2-sol
- Context window: 1.05M tokens (128K output)
- Price (Astra / Sol): $8 / $40 and $2 / $10 per 1M tokens
- Input / output: Text and images in, text out
- Availability: ChatGPT Work + Codex (Plus, Pro, Enterprise, Edu), API
- Safety rating: Critical for cyber; High for bio/chem

## FAQ

### How much does GPT-6.2 Astra cost compared to GPT-6 Astra?

GPT-6.2 Astra costs $8 per 1M input tokens, $0.40 per 1M cached input tokens and $40 per 1M output tokens. OpenAI says the input and output rates are 20% lower than GPT-6 Astra's, and cached input gets a 60% discount. Because GPT-6.2 Astra also writes fewer tokens, OpenAI reports many tasks cost under a third of what GPT-6 Astra charged.

### Should I use GPT-6.2 Sol or GPT-6.2 Astra for coding?

OpenAI recommends GPT-6.2 Sol for everyday coding and writing and GPT-6.2 Astra for the hardest projects. On Cognition's FrontierCode 2.0, which checks whether code changes meet maintainers' merge standards, GPT-6.2 Sol comes within 0.5 points of GPT-6.2 Astra at about one-fifth of the cost, with both at max reasoning effort.

### Who gets Ultrafast mode with GPT-6.2?

Pro 500 subscribers can use Ultrafast with both GPT-6.2 Astra and GPT-6.2 Sol. Pro 200 subscribers get Ultrafast for the first time, starting with GPT-6.2 Sol. Enterprise and Edu users get it once a workspace admin turns it on, and the OpenAI API offers Ultrafast for both models at launch. 9to5Mac reports Ultrafast runs up to 4.5 times faster but uses more of the allowance.

### How does GPT-6.2 Astra compare to Claude Fable 5.2?

OpenAI reports that GPT-6.2 Astra scores 5.5 points above Claude Fable 5.2 on Agents' Last Exam at about 5% of the cost, and more than doubles Fable 5.2's pass rate on Harvey LAB-AA v1.1 legal work. On Artificial Analysis's AA-3D human-preference test, GPT-6.2 Astra has 1,318 Elo against 1,271 for Claude Fable 5.2. These are OpenAI's selected comparisons.

### Why does GPT-6.2 refuse some security requests?

OpenAI rates GPT-6.2 Astra and GPT-6.2 Sol Critical for cybersecurity under its Preparedness Framework, so both launched with tighter safeguards for higher-risk cyber help, which also block some legitimate security research. Approved defenders can use OpenAI's Daybreak program for fewer refusals, and OpenAI says it plans to bring GPT-6.2 models to Daybreak.

## Links

- [System Card](https://deploymentsafety.openai.com/gpt-6-2-astra-sol)
- [Astra API Docs](https://developers.openai.com/api/docs/models/gpt-6.2-astra)
- [Sol API Docs](https://developers.openai.com/api/docs/models/gpt-6.2-sol)
- [9to5Mac Coverage](https://9to5mac.com/2026/11/17/openai-upgrades-chatgpt-and-codex-with-new-gpt-6-2-astra-and-sol-models/)
- [HN Discussion](https://news.ycombinator.com/item?id=50535230)
