Run effectively in production

Prepare your workflow for production by cutting unnecessary context and using prompt caching to reduce costs.

Match the model to your workload

Choose GPT-6 Astra for complex reasoning, GPT-6.1 Sol for coding, and GPT-6 Luna for repetitive tasks to balance cost and performance.

Adjust your prompts and skills

Define clear assignments, output requirements, and decision boundaries to guide the model effectively.

“Models have gotten much better at understanding nuance and ambiguity, so overly specific guidance can now hinder results where it previously helped.”

Optimize long-running tasks

Use steering, asynchronous tools, and delegation to manage complex work without delays.

Caching and compaction

Cached input tokens cost up to 95% less than uncached tokens, reducing costs for recurring work.

Model pricing and speed

GPT-6 Astra offers high intelligence for complex tasks, while GPT-6.1 Sol provides fast coding and research capabilities.

Reasoning levels

Adjust reasoning effort from low to extra high to match task complexity without breaking cache.

Speed settings

Use Fast mode for chat apps and Ultrafast for rapid coding iterations to balance speed and cost.

Real-world examples

Harvey, Cognition, Hex, and Invideo use GPT-6 Astra to improve drafts, testing, dashboards, and video editing.

Computer use

GPT-6 models interact with websites and desktop apps to fix bugs and test code directly.

From testing to production

Teams use GPT-6 Astra to create better drafts, test software, and build interactive dashboards with higher success rates.

Cost management

GPT-6 Astra costs $10 input and $50 output with $1 cached input, balancing cost and performance.

What happens next

Teams will use GPT-6 models to streamline workflows and reduce costs through better task management.

GPT-6 family

GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna offer different capabilities for complex reasoning, coding, and repetitive tasks.