Supported models

Load one model per session. LLMs unlock inspection, simulation, and evals; embedding models unlock the embed analysis suite. Run aquin list model anytime for IDs on your machine.

More model compatibility is on the way — including the ability to bring your own checkpoints with SAE training support.

LLMs

Inspection, simulation, evals, and checkpoint analysis.

• GPT-2 SmallLLMgpt2-small
• Pythia 70MLLMpythia-70m
• Pythia 2.8BLLMpythia-2.8b
• Llama 3.2 1BLLMllama-3.2-1b
• Sarvam 30BLLMsarvam-30b

Embedding models

Layer drift, retrieval, embed simulation, and embedding evals. See Inspection (embedding).

• all-MiniLM-L6Embeddingall-minilm-l6
• GTE SmallEmbeddinggte-small
• BGE Small ENEmbeddingbge-small-en

SAEs & checkpoints

• One model per session. Start a new session to switch.
• Feature-level tools need a public SAE: aquin list sae, then aquin load sae <model-l{n}>.
• Fine-tuned weights stay on disk — use --checkpoint while the base model id stays loaded.
• SAE catalog: aquin list sae · Checkpoint tools