Granite Guardian vs Mistral Moderation API

Granite Guardian

One ungated Apache 2.0 model judges harm, jailbreaks, RAG groundedness, function-call errors and custom criteria, with signed weights and published evaluation code. It is trained and tested on English only, each call checks one criterion, the 4.1 prompt format differs from 3.x, and IBM's watsonx.ai lists only the deprecated 3.0 model.

Mistral Moderation API

Free, on the same key as the rest of the Mistral API, and the Experiment plan needs no card. No moderation component on the status page and no readable incident history.

Granite Guardian

1. Append the <guardian> block as the final user message, with the mode line, ### Criteria: and ### Scoring Schema:. Copy the strings from the model card, because no package builds them
2. Use the no-think instruction for gating and parse <score>. Think mode writes a reasoning trace first, and the card's examples allow up to 2,048 output tokens
3. Treat yes as the criterion being met, which for built-in criteria means the risk is present. Treat a missing <score> tag as a failed check
4. Pass retrieved text through documents= and tool schemas through available_tools= in apply_chat_template, not inside the message text
5. Under Ollama, set num_ctx in the request options. IBM's docs say the default context is short and long requests are truncated

Mistral Moderation API

1. Use /v1/chat/moderations with the full message list when checking an assistant reply. The raw endpoint has no context
2. Read category_scores and set your own threshold per category. The booleans use Mistral's cut-offs
3. Pin mistral-moderation-2603. The 2411 model was retired on 31 March 2026
4. Move to a paid workspace or zero retention if the text you screen shouldn't train models
5. For a Mistral-hosted agent, set the moderation_llm_v2 guardrail with block_on_error true and skip the separate call