An X post from @lyraxana, an outside account not run by Google, claims Google is working on a "Decision" model and a Gemini API method called createDecision, but the Gemini API release notes through October 13, 2026 list neither. The post is a single line and comes with no screenshot, code diff or documentation link. The Gemini API release notes instead cover Gemini 4 Argon, Nano Banana 2.1, text-to-speech, Live and Antigravity changes.
What Google's pages do show
Google does publish decision-style tooling, just not as a Gemini API method. The MediaPipe Decision Maker guide on developers.google.com describes an on-device task that "evaluates structured decision questions" in a single forward pass, without generating text. It is built on three primitives:
ChoiceQuestion: picks one of K options and returns probabilities for each.
BooleanQuestion: returns a probability and a true/false value against a threshold (default 0.5).
ScoreQuestion: grades an ordered rubric and returns an expected score.
It runs on lightweight backends such as EmbeddingGemma, Laya, GLiNER2.5-Decide and Gemma 4 E2B/E4B, and it can parse TypeSafe Jev JSON payloads and OpenAI json_schema definitions directly.
The same guide's comparison table lists the TypeSafe Jev API as a server-side option, with "Server / Cloud / Python Harness (Vertex AI, Gemini API)" as its execution target. That describes where a Jev harness can run, not a Gemini API feature. A Google codelab on the same idea describes Jev as TypeSafe AI's "System One decision model": a hosted model that returns typed answers with calibrated probabilities in roughly 70 to 500 ms, priced at $0.042 per million input tokens with no output charge. The codelab pairs it with DiffusionGemma, Google's open-weights alternative, which can be self-hosted behind the same API shape. The codelab page is unverified, so treat its latency and price figures, which are TypeSafe's, as secondary; the MediaPipe guide is the better source for how Jev fits with the Gemini API.
Our read
Decision workloads appear as an on-device MediaPipe task and as TypeSafe's Jev API and SDK shape, not as a native Gemini model or method. Nothing Google has published in the pages we read says a first-party createDecision is coming.
If you build on the Gemini API, the practical check is the release notes and the API reference.





0 comments
No approved comments yet. You can start the conversation.
Leave a comment