Adding models¶

Models can be registered three ways — direct API call, env var, or YAML.

The model registry is only used for /api/tags discovery (so your model names appear in client dropdowns). You can call any prefix/model-id without registering it, as long as the prefix is configured.

Method 1 — Direct API call (no registration)¶

Any <prefix>/<model-id> works as long as the prefix is configured:

curl http://localhost:11434/api/chat -d '{
  "model": "openai/gpt-5-preview",
  "messages": [{"role": "user", "content": "Hi"}]
}'

The model just won't appear in GET /api/tags — useful for ad-hoc model testing without restart.

Method 2 — `EXTRA_MODELS` env var¶

Comma-separated list, no editing of any file:

export EXTRA_MODELS="openai/gpt-5,gemini/gemini-3-pro,kimi/kimi-k2.7"
ollama-proxy

These appear in /api/tags and ollama-proxy list-models.

Method 3 — YAML config¶

# ollama-proxy.yaml
models:
  - openai/gpt-5
  - gemini/gemini-3-pro
  - myvllm/meta-llama/Llama-3.1-70B

ollama-proxy --config ollama-proxy.yaml

Default registry¶

The proxy ships with a small default list of well-known models. See the source for the current list.

To get a quick view:

ollama-proxy list-models

Precedence¶

Models from all three sources are merged. Duplicates are skipped:

default registry  →  EXTRA_MODELS  →  YAML config

The order doesn't matter for routing — the registry is just for discovery.