Gpt-6 astra shipped with 1.1m tokens of context and a computer-use eval on day one. that is not a…
gpt-6 astra shipped with 1.1m tokens of context and a computer-use eval on day one. that is not a model upgrade — that is a model that now reads your whole monorepo and clicks your buttons. the bench that mattered was never frontiermath, it was can it finish your pr without you watching. last week the answer was no.
today, ask again.
Context
OpenAI's model page for GPT-6 Astra lists a 1,050,000-token context window and 128,000 max output tokens. OpenAI's launch post names computer use as a headline area and reports OSWorld 2.0 (v2026.08.08, offline set, partial score) of 72.6% at about 40 minutes per task, against 65.7% at about 75 minutes for GPT-5.6 Sol. It also reports Agents' Last Exam at 59.3% and a Codex harness update with 1.9x faster Mind2Web task completion.
The context window is exactly 1,050,000 tokens, so 1.1M rounds it up. The OSWorld figures are OpenAI's own, not independent, and the Claude comparison settings are footnoted as differing. A computer-use result on day one is supported. Whether it can finish a pull request unattended, and what the answer was last week, have no inspected evidence. The same launch post also reports FrontierMath Tier 4, so the claim that frontier math was never the bench that mattered is the author's view, not a finding.
Related work
- Path to Astra (OpenAI, 1 Sep 2026) ↗OpenAI's earlier post on the release.
- GPT-6 Astra system card (OpenAI) ↗The system card.
Watch next
- Third-party evaluations of long-running Codex and Astra tasks.
Sources
- GPT-6 Astra model page (OpenAI API docs)developers.openai.com
- Introducing GPT-6 Astra (OpenAI)openai.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 6 September 2026 at 02:03 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →