Latenode

GPT-6 Astra: $10/$50 API Pricing With a Still-Gated Rollout

Treat Astra as a gated, high-cost agent route: stay under the 272K cliff, plan for safety stops, and compare cost per completed job—not launch scorecards.

Editorial illustration for GPT-6 Astra: a gated high-cost agent route with pricing cliff and safety stop cues

OpenAI published GPT-6 Astra with standard API rates of $10 per million input tokens, $50 per million output, and $1 cached—plus full-request surcharges above 272,000 input tokens. Access still looks staged: OpenAI describes a phased rollout, and independent trackers date the limited-preview release to 2026-09-03, so broad availability cannot be confirmed from the crawl.

Per the OpenAI announcement, Astra targets computer-use and longer agent runs, with browser-style computer use, search, code interpreter, shell, MCP, and structured outputs. Production monitoring can also pause or stop tasks on potential unauthorized behavior—a real constraint for unattended workflows.

Independent analysis places Astra near GPT-5.6 Sol on a broader composite index: treat it as a premium specialist route, not proof of a universal intelligence jump.

When models and APIs shift keep the workflow running

Connect the models and apps from stories like this in one scenario. If a vendor changes routing, pricing, or availability, you update the workflow — you don't start over.

Start Free

Free forever plan. No credit card required.