Supported base models

Hugging Face base models Loops accepts, with dtype and sequence-length limits.

Each row below is a Hugging Face repo ID you can pass as base_model when starting a Loops run. The table also lists the inference, trainer, and LoRA dtypes Baseten provisions for the model, plus the maximum supported sequence length. Baseten adds rows as new models are validated end to end.

Models

Model	Max sequence length
Qwen/Qwen3.6-35B-A3B	131,072
Qwen/Qwen3.6-27B	131,072
Qwen/Qwen3.5-9B	131,072
Qwen/Qwen3.5-4B	131,072
Qwen/Qwen3.5-2B	131,072
Qwen/Qwen3.5-0.8B	131,072
Qwen/Qwen3.5-122B-A10B	Contact support
Qwen/Qwen3.5-397B-A17B	Contact support
moonshotai/Kimi-K2.6	Contact support
Qwen/Qwen3-30B-Instruct-2507	131,072
deepseek-ai/DeepSeek-V4-Pro	Contact support
deepseek-ai/DeepSeek-V4-Flash	Contact support
zai-org/GLM-5.1	Contact support
MiniMaxAI/MiniMax-M2.7	Contact support

Dtypes

The trainer dtype is the precision used for forward, backward, and optimizer steps. The LoRA dtype is the precision of the adapter weights. The inference dtype is the precision the paired sampling server uses to serve checkpoints.

Pass a model to Loops

Pass the table value verbatim as base_model through any of the following entry points:

• The Python SDK, via tinker.ServiceClient.create_lora_training_client(base_model=...). See the Loops quickstart.
• The HTTP API, via POST /v1/loops/runs.
• The CLI, via truss loops push <base_model>, which provisions a session, run, and paired sampler in one call.

The minimal HTTP call provisions a run and its paired sampler against an existing session. Replace sess_xyz789 with the session.id returned by POST /v1/loops/sessions:

curl --request POST \
  --url https://api.baseten.co/v1/loops/runs \
  --header "Authorization: Api-Key $BASETEN_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
    "session_id": "sess_xyz789",
    "base_model": "Qwen/Qwen3.5-9B",
    "max_seq_len": 131072
  }'

For the full request body, response shape, and remaining routes, see the Loops API reference.

Request a model

To request a base model that isn’t listed, contact support.