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

A roundup of OpenAI DevDay: GPT-6.1 Sol, the always-on agent Dots, ChatGPT Space, a $500 Pro 500 plan, Ultrafast mode and the Decisions API.

A satirical thread poking fun at OpenAI's launch event, including dots, ChatGPT Space, the Decisions API, GPT-6.1 Sol and a $500 plan.

On the author's own detection benchmark, GPT-6.1 Sol ranks second, 2 points behind GPT-6 Astra, while costing 78% less.

A week on GPT-6.1 Sol at MAX used only 2% of the weekly quota and 2.8M tokens, but the model is extremely slow with mediocre quality, so the author suggests Sol as implementor and Astra as orchestrator.

Sol scores 75.2% on DeepSWE, slightly above GPT-6 Astra's 74.1%, at nearly 7x lower cost per task.

The author lists three reasons for user dissatisfaction: no direct answer to Opus 5.5 in raw performance, reduced usage limits, and GPT-6.1 Sol launching only in Work and Codex with no Chat mode.

Federico Viticci shows Astra running a quick agentic task on Ultrafast, pulling two icons from the ChatGPT app for Mac in 30 seconds.

A Japanese user summarizes DevDay: dots runs on GPT-6 Astra with a cloud computer, alongside a cheaper Sol and an Ultrafast speed tier.

A rundown of the Dots always-on agent, the Space collaboration workspace, GPT-6.1 Sol at one-fifth the price, Ultrafast speedups, a $500 Pro tier, and Codex limit resets.

Japanese outlet Impress reports that GPT-6.1 Sol approaches Astra-level intelligence at about one-fifth the price, Astra Ultrafast has started running, and a new $500-per-month Pro tier was added.

The author sarcastically claims Claude Sonnet 5.5 took only 7 minutes and has better output throughput, so it 'wins' over GPT-6 Astra.

A Spanish user argues OpenAI's new agent dots reuses a setup already seen in Grok Bot and Meta Muse, noting that the dot.com domain redirects to a competitor.