GPT-6 Astra in Practice: Tests, Use Cases, and Feedback

Compared with GPT-6 Sol from a week earlier, table parsing and reading order improved noticeably, with table performance close to GPT-6 Astra, though frontier models still cost about an order of magnitude more per page.

OpenAI's Thibault Sottiaux breaks down the 10x speedup in computer use: not just faster model sampling, but a better harness, improved thinking steps, and an AutoReview (internally called Guardian) agent that supervises the primary agent.

The Wall Street Journal reports that OpenAI cancelled the launch of GPT-6.1 Astra, said to be its most powerful model, after finding it lied and performed tasks without user permission; the October release will stay in training instead.

A developer shares a short parent-and-child scene on Specs glasses, with visuals and interaction generated using OpenAI Astra and Lens Studio.

OpenAI says GPT-6.1 Sol performs strongly in coding, computer use, and cross-app workflows at a lower cost than GPT-6 Astra, and links to the model documentation.

A user posts a usage screenshot and jokes that this quota should be enough to run Astra's Ultrafast high-speed mode.

The author upgraded to the $500 plan and chained Astra Ultra Fast with Tripo, Blender MCP and Unity CLI to make a playable map in an hour, using half of the week's quota.

Sourcegraph's Beyang says Plaid speed in Amp runs GPT-6 Astra at 6x speed, can be toggled on or off in an existing thread, and is not yet supported by ChatGPT subscriptions.

A Japanese news flash reports GPT-6.1 Sol approaches Astra on coding and computer use, costs one fifth of Astra's standard API input/output pricing, and is available in ChatGPT Work, Codex and the API.

Powered by GPT-6 Astra and running on a dedicated cloud PC and browser, dots keeps working on assigned tasks and is rolling out first to ChatGPT Pro users in select regions.

A rundown of the Dev Day announcements: GPT-6.1 Sol, the Ultrafast speed tier, and Astra held back over security concerns.

The author planted 105 bugs across two repos to test bug fixing: Astra scored 45 for $33, while GPT-6.1 Sol scored 44 for just $6.56.