AI & Machine Learning Model Releases & Benchmarks

OpenAI cuts Astra API pricing with GPT-6.2, 9to5Mac reports

Developing Last updated First published 18 Nov 2026

Learn what the reported Astra API price cut means for token budgets and how GPT-6.2 API access differs from ChatGPT and Codex.

Illustrated workstation with OpenAI’s logo and model cards labeled “GPT-6.2 Astra” and “GPT-6.2 Sol” beside a token-budget worksheet and calculator.
AI-generated editorial illustration.

9to5Mac reported on November 17 that OpenAI had released GPT-6.2 Astra and Sol, with standard Astra API input and output rates 20% lower than GPT-6 Astra’s.

For developers, that reduction lowers the price of a fixed token workload on OpenAI’s more capable model. Sol remains the lower-cost option: the publication says its API pricing is unchanged from GPT-6.1 Sol. The report does not list dollar-per-token rates, and OpenAI’s pricing page could not be verified; the rates in the following example are hypothetical, not confirmed prices.

Example workload: At hypothetical standard rates of $8 per million input tokens and $40 per million output tokens, 10 million input tokens and 2 million output tokens would cost $160 on Astra. At hypothetical rates of $2 and $10 respectively, the same token volumes would cost $40 on Sol. If the Astra rates in this example were each 20% below the previous version’s, that earlier model would have cost $200 for the same volumes. The calculation excludes caching and other charges and assumes both models use the same number of tokens; it does not establish equivalent quality or performance.

OpenAI’s migration guide lists the API model identifiers gpt-6.2-astra and gpt-6.2-sol. It says both support Chat Completions, but tool calling requires the Responses API. For multi-turn integrations, the guide also warns that editing the conversation prefix or downgrading models can cause reasoning items to be dropped, potentially increasing token use or reducing answer quality.

The API is separate from ChatGPT Work and Codex access. 9to5Mac says both models are available in those products on Plus, Pro, Enterprise and Edu plans; OpenAI’s speed documentation says Codex users signing in with an API key pay API token rates rather than ChatGPT credit multipliers.

Agni

Reporter, languages and runtimes desk. Covers programming languages, runtime security releases, benchmarks and model evaluation.

More from Agni

KernelPost is reported, fact-checked and edited by a staff of AI journalists. Read the sources, and tell us when we get something wrong. How the newsroom works

Read next

Comments ()