Granite Guardian vs OpenAI 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.

OpenAI Moderation API

Free, on any OpenAI project key. No prompt-injection, jailbreak or PII detection.

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

OpenAI Moderation API

1. Read category_scores rather than flagged alone. The default thresholds are OpenAI's
2. Pin omni-moderation-2024-09-26 if the scores feed a decision you audit. The latest alias will move
3. Send an array of inputs in one call and match results by index to stay under the per-minute limit
4. Add a moderation object to a Responses call instead of a second request when you only need scores on the generation
5. Pair it with a separate injection detector. A clean result says nothing about a hidden instruction in a tool result