AIBase
A practical first workflow
- Define a representative writing / coding task and its acceptance criteria.
- Pin the exact model identifier and confirm hosted api availability for your environment.
- Run a small evaluation that records correctness, latency, input/output usage and review time.
- Compare the result with another suitable model before scaling usage.
Before you commit
- Workflow fit: High-volume tasks where cost matters and outputs can be checked.
- Limits and trade-offs: Evaluate error rates and retries before choosing on token price alone.
- Full cost: Include input/output usage, hosting where relevant, retries, tools and evaluation time.
- Data and permissions: Read the current terms for the chosen account or deployment. Confirm who can access uploaded material and which actions require review.
Common questions
Where should I start with GPT-6 Luna?
High-volume tasks where cost matters and outputs can be checked. Start with one small task that you can check manually, then use the workflow above to evaluate the result.
What does this profile verify?
It summarises provider documentation and explains a practical evaluation approach. Evaluate error rates and retries before choosing on token price alone. It does not establish an independent performance score.
How do I compare the alternatives?
Use the same inputs, definition of done and review process for each option. Compare total cost and output quality together. Build a comparison board or find a shortlist by task.
Published API pricing
Input / 1M tokens$0.1
Output / 1M tokens$0.5
Source & verification
This profile summarises provider documentation. It is not a hands-on product review or an independent benchmark.
Developers.openai.com · Official source ↗Found a change? Send a correction with the source URL.
