Google's new decision endpoint returns calibrated probabilities instead of prose, and the free tier covers 10,000 calls a day.
Gemini Decision 1 (gemini-decision-1) is built for bounded choices: yes/no checks, picking from up to 255 options, or scoring against a rubric. One call can carry up to 1,000 questions over the same input — text, images, audio, video or PDFs, up to 262,144 tokens.
The catch: free-tier prompts are used to improve Google's products, the 50,000-tokens-a-minute ceiling binds before the request cap on long inputs, and there is no Batch API or context caching.
How to claim: 1) Create a Gemini API key in Google AI Studio → 2) Install google-genai SDK 3.2 or later (pip install -U google-genai) → 3) Call client.decisions.create with model gemini-decision-1.
Deadline: Ongoing.
What changed
Google released the Decisions API and Gemini Decision 1 on Nov 11, 2026. The free tier lists 200 requests a minute, 50,000 tokens a minute and 10,000 requests a day. Paid usage is $0.10 per 1M input tokens with zero output tokens billed, at 150–250 ms latency.
What it means for you
10,000 calls a day covers a routing or policy check on every message in a busy support queue, without a bill. On the paid tier, 10M input tokens of the same work costs $1 at $0.10 per 1M; a 2,000-token prompt costs $0.0002. Because output tokens are always zero, the price depends only on what you send in.
Anyone with a Google account can use the free tier: no card, no billing account, no application. Google uses free-tier prompts to improve its products; it says paid-tier prompts are not used. Limits: 10,000 requests a day, 50,000 tokens a minute, 200 requests a minute, 1,000 questions per request. The offer page: /offer/google-gemini-developer-api-free-tier.
Related free offers
- Google Gemini Developer API Free Tier — Free Gemini API tier with rate-limited access to Flash models and Gemini 2.5 Pro
- Cohere Free Trial API — Free rate-limited Cohere Trial API key for non-commercial prototyping
- Groq Free Tier — Free rate-limited Groq API access to hosted open models, no card