Compare the work
Start with the task you need to run, then compare context, response behavior, and price fields for the routes that fit.
OpenAI-compatible models, with fewer surprises in the request path.
The client shape is familiar. The important details are still explicit: model ID, context, account group, rate date, response status, and request record.
Start with the request your team needs.
A small smoke test proves the route works. The next request should resemble the workload you actually plan to run.
One representative payload, one saved resultCompare the work
Switch in one field
The OpenAI-compatible request shape lets you test another model without rewriting your client integration.
Keep the evidence
A model comparison is more useful when it includes the prompt shape, status, usage, latency, and charge.
Check the details that decide a successful call.
The client shape is only the beginning. These checks keep a working request understandable after it leaves your terminal.
Route, account, response, recordModel directory
Use the current directory as the source for exact IDs. Retired or mistyped IDs return a different failure than a bad key.
Context and output
A documented context window is only one boundary. Reserve output and account for timeouts and rate policy.
Rate date
Prices change. Use a dated rate card and verify the effective group before you compare two workloads.
Run one small request before you scale it.
Keep the key server side. Copy the request, replace the model only with an ID from the current directory, and inspect the response.
OpenAI-compatible chat completionsCopy-ready request
curl https://aiwave.live/v1/chat/completions \
-H "Authorization: Bearer $AIWAVE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"deepseek-v4-pro","messages":[{"role":"user","content":"Say hello in one sentence."}],"max_tokens":32}'Before production
Keep the first result explainable.
Save the request ID, model ID, usage fields, status, and charge with the application record.
- Check the key without printing its value.
- Use a current model ID from the directory.
- Set a client timeout and bound retries.
- Compare the response record with the dated rate.
Questions developers ask before integration.
What does OpenAI-compatible mean here?
The chat completions request shape and client pattern are compatible. You still choose a model ID from the current directory.
Can I compare models before making a large request?
Yes. Use a small representative prompt, record the response, then compare usage and charge against the dated rate.
Where are the model IDs?
The public model directory is the current source for IDs and links to the pricing and documentation pages.
Run a request you can inspect afterward.
Keep the request small, confirm the route, then move to the workload that matters.