Sulat.com
AI models
Vercel AI Gateway logo

Model details

GPT-6 Astra

GPT-6 Astra is OpenAI’s reasoning-focused model for difficult end-to-end work, combining advances in pre-training, reinforcement learning, and alignment. Rather than serving only as a conversational model, it is designed to plan, use software, revise intermediate work, and turn open-ended goals into polished documents, analyses, code, or other professional artifacts. Its published positioning centers on computer use, browsing, software engineering, science, cybersecurity, and knowledge work.

Astra’s practical advantage is strongest in sustained, tool-using workflows. OpenAI reports state-of-the-art results in several of these areas, including 97.6% on FrontierMath Tier 4 v2, 72.6% on OSWorld 2.0, and 100% on ExploitBench, while also showing strong results on Terminal-Bench Science, BrowseComp, GPQA Diamond, and very-long-context retrieval. Independent coverage still finds it less dominant on some broad aggregate measures, so the best fit is high-stakes agentic work where fewer retries, stronger execution, and deeper context matter more than a universal benchmark win.

Vercel AI Gatewayopenai/gpt-6-astragpt-astra

Quick Info

Powered by
Provider
Vercel AI Gateway
Model key
openai/gpt-6-astra
Release date
Sep 4, 2026
Last updated
Sep 4, 2026
Knowledge cutoff
2026-04-30
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$10.00
Output token cost
$50.00

Limits

Input tokens
922,000 tokens
Output tokens
128,000 tokens
Context window
1,050,000 tokens

Transparent token rates

Compare GPT-6 Astra pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about GPT-6 Astra

Ofox

CoverageBenchmark

OpenAI released GPT-6 Astra on September 3, 2026, framing it as "the world's most intelligent and aligned model" and the product of its largest training run to date, executed on more than 100,000 GPUs at the Stargate site in Texas. It is also the first release where earlier OpenAI models supervised the training of the Astra also becomes the first model to hit the "Critical" cybersecurity threshold under OpenAI's Preparedness Framework, which is why advanced cyber capabilities are being gated behind the Daybreak application-based program and the rollout is intentionally phased. Vellum notes that not every benchmark favors OpenAI: Met

Ofox

Coverage

OpenAI announced the phased rollout of GPT-6 Astra on September 3, 2026, with CEO Sam Altman calling it a "new capability level" that has changed his workflows and predicting "a boom of entrepreneurship, of creativity, of economic growth, of scientific discovery." Distribution channels for Astra include ChatGPT Plus, P Companies participating in OpenAI's application-based Daybreak cybersecurity program are first in line for access, because Astra is OpenAI's first model to reach the "Critical" internal cybersecurity threshold disclosed earlier in the week, with those advanced cyber capabilities explicitly limited. CNBC notes that two

Videos about GPT-6 Astra

More models around GPT-6 Astra