GPT-6 Astra is officially here. OpenAI launched its latest flagship model on September 3, 2026, after slowing down its release because of concerns around advanced cybersecurity capabilities. The new model focuses heavily on software engineering, computer use, reasoning and complex technical tasks.
The official API model ID is gpt-6-astra, and it comes with a huge 1,050,000-token context window. However, this extra capability also comes with a much higher price than the previous GPT-5.6 Sol model.
GPT-6 Astra Release Date and Rollout
OpenAI released GPT-6 Astra on September 3, initially giving access to a limited group of organizations through its trusted-access programs.
The rollout is happening in stages. ChatGPT Plus, Pro, Business and Enterprise users are receiving access over time, while API availability is also expanding. Enterprise access is disabled by default until an administrator enables it.
Inside ChatGPT, Astra is also connected with the GPT-6 Pro name on supported paid plans.
GPT-6 Astra Pricing
One of the biggest changes is the price. GPT-6 Astra costs $10 per million input tokens and $50 per million output tokens through the API.
| Feature | GPT-6 Astra | GPT-5.6 Sol |
|---|---|---|
| Input / 1M tokens | $10 | $4 |
| Output / 1M tokens | $50 | $20 |
| Cached input | $1 | $0.40 |
| Context window | 1.05M | — |
| Maximum output | 128K | — |
Cached input is cheaper, while batch and flex processing are available at half the standard price. Fast mode costs twice the normal rate.
For businesses handling large volumes of requests, this makes model routing important. Expensive models can be reserved for difficult tasks while cheaper models handle everyday work.
GPT-6 Astra Features and Specifications
The model supports text and image input and can produce text output. It offers reasoning levels ranging from low to max and supports tools such as web search, file search, code execution, hosted shell, computer use and MCP.
Its knowledge cutoff is April 30, 2026.
The main focus is not simple conversation. OpenAI positions GPT-6 Astra as a model designed for software development and computer-use tasks.
GPT-6 Astra Benchmarks
OpenAI reports major improvements over GPT-5.6 Sol on several internal benchmarks.
| Benchmark | GPT-6 Astra | GPT-5.6 Sol |
|---|---|---|
| Terminal-Bench 4.0 | 57.9% | 37.3% |
| OSWorld 2.0 | 72.6% | 65.7% |
| FrontierMath Tier 4 | 97.6% | 83.0% |
| GPQA Diamond | 96.0% | 94.6% |
| ExploitBench | 100% | 78.5% |
These figures come from OpenAI’s own testing, so independent testing is still important before making major purchasing or development decisions.
The Cybersecurity Restriction
Cybersecurity is one of the most notable parts of the GPT-6 Astra launch. OpenAI says the model reached its Critical cybersecurity threshold during testing.
The public version has restrictions around advanced cyber tasks, including creating proof-of-concept exploits. More flexible capabilities are available to vetted organizations through OpenAI’s Daybreak program.
This security layer was also a major reason behind the delayed rollout.
What Developers Should Know
GPT-6 Astra is powerful, but its $50-per-million-token output price makes it expensive for routine workloads. A practical approach is to use it for complex coding, agentic workflows and difficult technical tasks while routing simpler requests to lower-cost models.
Using an OpenAI-compatible API can also make it easier to change models later without rebuilding an entire application.
Final Thoughts
GPT-6 Astra marks a major step toward AI systems that can handle complex software and computer-based tasks. Its large context window and strong benchmark results make it interesting for developers, while its pricing and cybersecurity restrictions will shape how businesses actually use it.
For most teams, the key question is not simply whether GPT-6 Astra is powerful. It is whether the extra capability is worth the additional cost for their specific workload.
