| Model | Slug | Image | Reasoning |
|---|---|---|---|
valar-auto-balanced | |||
moonshotai/Kimi-K3-fast | |||
zai-org/GLM-5.3-fast | |||
deepseek-ai/DeepSeek-V4.1-Flash-1017-fast | |||
Qwen/Qwen3.5-27B | |||
qwen/qwen3.8-max | |||
joey-1 | |||
elephant-1 |
Reasoning models and output caps
max_output_tokens caps reasoning tokens plus the visible answer, not the answer alone. On a model marked Reasoning above, most of that budget goes to thinking you never see: often a thousand reasoning tokens behind a two-sentence answer.
So a cap sized for the answer stops generation mid-thought, and the response settles incomplete:
incomplete with an answer that is just as complete as the rest. The only difference is how long the model thought.
Branch on incomplete_details.reason and read the output instead:
Size the cap for reasoning plus answer
Send a handful of representative requests, readusage.output_tokens_details.reasoning_tokens off the results, and set the cap to that plus the answer length you want. Omit max_output_tokens entirely when you don’t need a hard ceiling.
Custom models
Beyond the catalog above, Valar can serve your own model weights. If you have a custom or fine-tuned open-weight model, we can host it on Valar’s inference stack and expose it through the same OpenAI-compatible API, completion windows, and billing as any catalog model. Reach out to your Valar contact to onboard a custom model.To confirm what your API key can reach in a given environment at runtime, call
GET /v1/models against that environment rather than relying on this table alone. Each model it returns carries its context window, capability flags, and the rate card your organization is billed at.