Supported Models

Blindference currently supports three inference models across cloud and local backends.

Model Reference

Model ID	Provider	Backend	VRAM / Requirements	Speed	Quality
`groq:llama-3.3-70b-versatile`	Groq	Cloud API	`GROQ_API_KEY` env var	Fast	High
`gemini:gemini-2.5-flash`	Google	Cloud API	`GOOGLE_API_KEY` env var	Fast	Medium-High
`facebook/opt-125m`	Local	vLLM	0.5GB+ VRAM, `vllm` package	Variable	Dev/Testing

When a job arrives, the node queries backends in registration order:

The first backend that (a) is available and (b) advertises the requested model_id wins.

export GROQ_API_KEY="gsk_..."

Get your key at console.groq.com.

export GOOGLE_API_KEY="AI..."

Get your key at ai.google.dev.

pip install vllm

blindference-node models test --backend vllm --model facebook/opt-125m --prompt "Hello"

Backend	Determinism Method
vLLM	`temperature=0`, `seed=42`, `enforce_eager=True`
Groq	`temperature=0`, `seed=42` (API-native)
Gemini	`temperature=0` (best-effort)
Mock	SHA-256 of `(model_id, prompt)`

All cloud backends use a lightweight [seed_anchor:{hash}] prefix to reduce variance without restricting response creativity.

See Model Backends for the full pluggable backend system.