Microsoft-Decision-1 vs RightNow System Zero

Microsoft-Decision-1

Microsoft publishes $0.042 per million input tokens, with output free, and the model is callable through Microsoft Foundry and OpenRouter. It is in public preview. No licence, model-specific retention statement, rate limit or deprecation policy was found, and the Foundry route needs an Azure deployment and an Entra token. Latency and accuracy figures are Microsoft's claims.

RightNow System Zero

A well-documented decision API with an OpenAPI file, idempotent retries, a detailed error catalogue, streaming over WebSocket and published per-unit prices with USD 5 of starting credit. Standard context is 38,000 tokens; the 64k, 128k and 1M bands are held and video is a preview. The latency and benchmark figures are RightNow's own, and several of its pages disagree.

Microsoft-Decision-1

1. Use OpenRouter's Decisions method, not an OpenAI chat-completions SDK. OpenRouter says chat completions SDKs will not work with this model
2. On Foundry, send the request to the deployment's /providers/microsoft/v1/systemone path with a Microsoft Entra token for https://cognitiveservices.azure.com/.default, not an API key
3. Take the deployment name from the Foundry quickstart before the first call. Microsoft says to confirm the route and authentication header, and the pages reviewed do not give the name
4. Keep each request within OpenRouter's 32,768-token context and send only fixed options, since the model is not intended for open-ended generation
5. Measure latency and calibration on your own labelled cases before relying on Microsoft's latency and 'nine times out of 10' statements

RightNow System Zero

1. Send audio only as inline 16,000 Hz WAV. The FLAC, MP3, OGG and MP4 audio named on the blog are refused with media_unsupported.
2. Keep state under 38,000 tokens. Long-context requests return 404 long_context_not_available, which never succeeds on retry.
3. Reuse the same Idempotency-Key and body on retries, and stop when error.retryable or x-should-retry is false.
4. Check every entry in answers even on HTTP 200, since one question can be refused while the others are answered.
5. Treat probabilities as uncalibrated while calibrated is false, and pin system-zero-2.0.0 when answers must not move with the alias.