Granite Switch 4.1 vs Liquid d1

Granite Switch 4.1

One self-hosted Apache 2.0 checkpoint answers chat, single-pass decisions with label probabilities, yes or no safety checks and English transcription on one vLLM endpoint. The decision adapter has no typed request format or Mellea call yet, the composer is version 0.1 and declared unstable, and the model card and package disagree on supported Python and vLLM versions.

Liquid d1

The hosted API costs $0.02 per million input tokens for d1 and $0.042 for d1-ultra, and the d1-3B weights run locally through Transformers or llama.cpp. Liquid's terms allow it to train on API content, and no status page, rate limits or error reference were found.

Granite Switch 4.1

1. Pass chat_template_kwargs: {"adapter_name": "decision"} or "guardian-core". Without it the request is ordinary chat on the base model
2. For a decision, send the state and the ### DECISION question as two user turns, request logprob_token_ids for the option labels, and set max_completion_tokens to 1
3. Use single-token option labels. Split more than 128 labels into chunks with the identical prompt and apply one softmax over all returned logprobs
4. Parse guardian-core answers as JSON, {"score": "yes"} or {"score": "no"}, where yes means the criterion is met
5. Add authentication before exposing the vLLM server. Neither vLLM nor the checkpoint requires a key by default

Liquid d1

1. POST to https://api.liquid.ai/decisions/v1/systemone with a liquid_ key as a Bearer header. This is not a chat-completions endpoint
2. Use d1, or d1-ultra for harder decisions at $0.042 against $0.02 per million input tokens. d1:free is no longer documented
3. Send images as Base64 data URLs in images, at most 8, with the whole JSON body under 4.5 MB. Remote image URLs are refused
4. Budget tokens per question. Each question is billed as its own prompt, with its text and every image counted again
5. Don't send confidential content to the hosted API, since the terms allow training on it. Run d1-3B locally for that data