Models and Model IDs

The Models page is the source of truth for public model IDs. Use it when calling the API, configuring tools, and troubleshooting model_not_found.

Good starter models

Need	Start with	Why
Low-cost testing	`kimi-k2.6`	Fast API checks, automation, and agent loops.
General app features	`kimi-k3`	Text generation, extraction, support, and lightweight reasoning.

Copy the exact model ID

Model IDs are case-sensitive. Slashes and suffixes matter.

Correct:

kimi-k2.6

Different IDs:

Kimi-K2.6
kimi-k2-6
moonshot/kimi-k2.6

Public model status

The website uses these model status labels:

Status	Meaning
`Available`	Publicly listed and callable through the API when your key scope allows it.
`Price pending`	Sale price display is still being confirmed. On today's public shelf this badge is informational: models marked Available remain callable even when a more precise list price is pending.
`Tiered`	Public model with complex or tiered pricing normalized for display.
`Unavailable` / `Temporarily unavailable`	Not currently offered for new public traffic.

Callable models on the public shelf appear in authenticated GET /v1/models. The website Models page is the browseable source of truth for IDs and badges.

Catalog visibility and API eligibility

A model that is listed as Available on Models is callable through the public API (subject to your API key scope and prepaid balance). Price pending alone does not mean "not callable." Treat Temporarily unavailable / Integration review labels as non-callable unless Models explicitly shows Available.

Kimi model notes

The following reviewed Kimi IDs are available:

Model ID	Best for	Request note
`kimi-k2.6`	General chat, coding, and agent workflows	Standard OpenAI-compatible chat parameters.
`kimi-k2.7-code`	Coding and repository tasks	If you send `temperature`, it must be `1`. Omitting it is also supported.
`kimi-k3`	Reasoning-heavy chat and coding	Use a sufficient `max_tokens` budget because reasoning tokens are generated before final text.

For Kimi K3, start with max_tokens: 512 or higher for normal answers. A very small value can be consumed entirely by reasoning and return no final text.

Key model scope

An API key can be:

unrestricted, which means it can use all current and future publicly available models;
limited to chat models;
limited to selected model IDs.

If a key is scoped too narrowly, a request can return model_not_found even when the model exists in the public catalog. Check the key's model scope before changing application code.

Troubleshooting `model_not_found`

Open Models.
Search for the model you want.
Copy the full model ID.
Confirm the key's model scope allows that model.
Retry the request.

Do not copy vendor aliases directly from another provider's documentation. NexoRouter public model IDs may differ from upstream aliases.