Set your requirements before sorting by price.
Choose a publisher or image-input filter, then enter one request’s uncached input and output tokens. Compare the listed context limits and API settings before choosing a model for your application.
Compare one uncached text request across published account rates. Cache, hosting and tools are excluded. Validate task quality separately.
Enter a workload to compare available model rates.
Codimixa rates checked: 2026-10-31T15:06:07.868Z. Publisher references checked: October 20, 2026.
Match the ranking to your workload.
A model with cheap input can suit long-document tasks. A model with cheap output can suit generation-heavy work. Cache rates matter only when your workload can use the relevant cache behavior.
| Workload | Compare first | Also inspect |
|---|---|---|
| Long documents | Uncached input cost | Context limits and answer quality |
| Long generated outputs | Output cost | Output limits and accepted results |
| Coding agents | Cost per completed task | Tools, retries and repeated context |
| Repeated prompts | Applicable cache rates | Measured cache use and write costs |
Compare official and Codimixa rates for the same model.
Selected models have rates from 0.3× their official API reference. Use the exact rate for each candidate when you compare them.
Give finalists the same task.
Use one small, repeatable task set. Record completion, total API cost, extra attempts and any manual repair. Keep the cheapest candidate that meets your acceptance criteria.
A FEW DETAILS
Before you continue
Is one LLM API always cheapest?
The result depends on your token mix, required features and completion rate. Change the workload and the cost ranking can change too.
Does this compare every API service?
The model comparison uses Codimixa’s available catalog and linked official API references. It is not an exhaustive ranking of every service on the market.
Find your lowest-cost viable model.
Start with the catalog filters, then estimate the same workload for your shortlist.
Browse models