Models

Retrieve information about available models.

Live IDs below were verified against the cluster on 2026-09-04 (vllm.lab.hq.solidrust.net and embeddings.lab.hq.solidrust.net).

List Models

GET /v1/models

Example Request

curl https://api.solidrust.ai/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

Chat completions are served by the vLLM chat deployment. Embeddings are a separate vLLM process — GET /v1/models on the chat base URL lists the chat alias. The embeddings model id is what /v1/embeddings expects.

Chat response (live)

{
  "object": "list",
  "data": [
    {
      "id": "vllm-primary",
      "object": "model",
      "owned_by": "vllm",
      "root": "google/gemma-4-12B-it-qat-w4a16-ct",
      "parent": null,
      "max_model_len": 16384
    }
  ]
}

Available Models

Chat Models

Model ID	Currently serving	Context	Best For
vllm-primary	Gemma 4 12B IT QAT (google/gemma-4-12B-it-qat-w4a16-ct)	16384	All chat tasks — use this alias

Use vllm-primary

Always use vllm-primary in your code. The alias is the public contract. The weights behind it change; your code should not.

There is no public qwen3-4b id. That model is gone.

Embedding Models

Model ID	Dimensions	Max input	Best For
Qwen/Qwen3-Embedding-0.6B	1024	32768 tokens	Semantic search, RAG

There is no public bge-m3 id. The embedding service lists Qwen/Qwen3-Embedding-0.6B only.

Model Selection

• Chat completions: vllm-primary
• Embeddings: Qwen/Qwen3-Embedding-0.6B

Failover

POST /v1/chat/completions goes Artemis → vLLM. LiteLLM is retired.

The agent endpoint (POST /v1/agent/chat) has a separate provider chain: local vllm-primary, then OpenRouter (openrouter/auto) if the data-layer agent health check says vLLM is down. That failover does not apply to plain /v1/chat/completions.