Google launches Gemini Decisions API with Gemini Decision 1
Original title: Google Launches Gemini Decisions API With Gemini Decision 1 at $0.10/M Tokens
The hosted model scores up to 1,000 closed-set questions per request at $0.10 per million input tokens, with no output charge.
Google released the Gemini Decisions API, powered by Gemini Decision 1 (gemini-decision-1), a multimodal model that returns calibrated probabilities instead of generated text. It supports yes/no, choice, and rubric scoring over a shared input of text, images, video, audio, and PDFs, with a 262,144-token context. The endpoint is live on the Gemini Developer API (v1beta) and needs google-genai SDK 3.2 or later; open weights and self-hosting are not offered. Google’s own table says it leads 7 of 9 accuracy benchmarks and has the lowest calibration error.
Unconfirmed: Accuracy, calibration, latency, and price comparisons are self-reported by Google.
- Google reports 92.2% on JevBench and 58.7% Macro-F1 on Amazon ESCI.
- Claimed end-to-end latency is 150-250 ms; median compute is 50-100 ms.
- Shared-suite cost is listed at $0.88 versus $4.14 for Gemini 3.5 Flash-Lite.
- It trails TypeSafe Jev 1.14 on When2Call, 79.9% versus 82.0%.
- Google says internal Notebook routing cut everyday-query latency by up to 40%.
- Vendors
- Products
- Gemini Decisions API
From the original article
8 min read at marktechpost.comGoogle launched the Gemini Decisions API, powered by Gemini Decision 1. It returns calibrated probabilities for up to 1,000 yes/no, choice and rubric questions over text, images, audio, video and PDFs, at $0. 10 per 1M input tokens.
By Michal Sutter · marktechpost.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.