The Daily Byte
Breaking
Welcome to The Daily Byte — real stories from across Pakistan.
Saturday, 21 November 2026 · Pakistan

The Daily Byte

Stories from Pakistan that deserve to be heard
Minor World

OpenAI Launches GPT-6.2 Astra and Sol Models

By The Daily Byte ·21 Nov 2026 ·3 min read
OpenAI headquarters office building in San Francisco

OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol, two new artificial intelligence models aimed at professional work, and cut the price of its flagship model by a fifth. The company announced both models on 17 November, and they are available now in ChatGPT Work, in its Codex coding tool and through the API.

Astra is the new frontier model, built for long professional workflows such as slides, spreadsheets, legal documents, scientific work and agentic coding. Sol is the lower-cost model for everyday technical and non-technical work. Both have a 1,050,000-token context window and a maximum output of 128,000 tokens.

Astra gets cheaper, Sol stays put

The headline change for developers is Astra's price. It now costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for GPT-6 Astra — a 20 percent cut on both sides. Cached input, the text the system has already processed and can reuse, falls further, from $1.00 to $0.40 per million tokens, a 60 percent reduction that matters most for agent workloads that resend long prompts again and again.

Sol keeps exactly the price of its predecessor, GPT-6.1 Sol, at $2 per million input tokens, $0.10 cached and $10 per million output tokens. The Sol upgrade therefore arrives with no change to the bill at standard rates. OpenAI says Astra sets new highs on professional and scientific evaluations at less than a third of GPT-6 Astra's cost per task.

Where you can use them

Paid ChatGPT subscribers on the Plus, Pro, Enterprise and Edu tiers can choose either model inside ChatGPT Work or Codex. Developers can reach them in the API under the IDs gpt-6.2-astra and gpt-6.2-sol. The release notes state that the rollout does not add either model to the regular Chat tab, which stays unchanged for most users. Ultrafast processing is available on both models, at three times the standard rate until 28 February 2027.

Safety ratings

Both models carry the biggest caveat in the release. Under OpenAI's Preparedness Framework, GPT-6.2 Astra and Sol are rated at the Critical level for cyber capability, and they ship with tighter guardrails on higher-risk security help, so some security research requests may be refused. The system card cites pre-deployment work by the UK AI Security Institute.

Why it matters

The launch lands in the same week that Google shipped Gemini 4 Neon, with both labs competing on cost per task rather than raw benchmark scores alone. Astra's new price matches Anthropic's Claude Fable 5.2 on headline token rates, but OpenAI argues the real cost of a finished job is far lower. For teams already running agents on the API, the cheaper cached input is the change most likely to show up in the monthly bill.