Google Gemini Decision 1 uses the Google Gemini Decisions API (model gemini-decision-1). Instead of writing a reply, it returns a typed decision about the text, images, PDFs, audio or video you submit, much faster than a chat model — ideal for high-volume, well-defined work such as content classification, request routing, policy checks, quality control and prioritization. Pick a question type on the panel first: Yes or no returns the probability that a condition is true, and you may describe what yes and no mean in the Criteria field as true|... and false|...; Choice picks one of the 2 to 255 options you supply and lists the probability of every option plus a confidence value; Score rates the input against the levels you list from lowest to highest and returns a probability-weighted score. Enter the condition or question in the Question field, list one option per line as value|description for choice and score questions, then paste the text to evaluate into the input box or upload images, PDFs, audio or video and send; uploaded files may total up to 64MB per request, of which PDFs up to 32MB. If the Question field is empty, the input text is used as the question and the uploaded files are the evidence. Results are shown as tables with probabilities and confidence. For written explanations or multi-turn conversations, use Gemini 3.6 Flash or another chat model instead. Prompt example 1: (Upload a support call recording, mp3; Panel: Yes or no; question "Does the customer clearly ask to cancel the subscription?") Evaluate this call. Prompt example 2: (Panel: Choice; question "Which department should handle this message?"; options accounts, tech support, shipping, other) My account was debited twice last month, please return the extra amount. Prompt example 3: (Upload a lease agreement, PDF; Panel: Yes or no; question "Does the lease include an early termination penalty?") Evaluate this lease. Prompt example 4: (Upload an unboxing video, mp4; Panel: Score; question "Damage to the product"; levels low to high: none, light scratches, clearly damaged, unusable) Rate the product in the video. Prompt example 5: (Panel: Yes or no; question "Is this post a personal attack?"; criteria: true|insults a person, false|criticizes the issue) Waited half an hour and still got the wrong order. Prompt example 6: (Upload a receipt photo, jpg; Panel: Choice; options meals, transport, lodging, office supplies, other) Which expense category does this receipt belong to? Prompt example 7: (Panel: Score; question "How polite is this email?"; levels low to high: rude, neutral, polite, very polite) Answer me today or I will file a complaint. Prompt example 8: (Panel: Choice; question "What does the user want?"; options one per line: search|look up, order|buy, support|get help, chat|small talk) Any train seats from Taipei to Kaohsiung tomorrow? Prompt example 9: (Upload a meeting recording, m4a; Panel: Choice; question "What was the final decision?"; options approved, rejected, postponed) Evaluate this recording. Prompt example 10: (Upload a poster design, png; Panel: Yes or no; question "Is blue the main color of this design?") Evaluate this design. Maximum context window (input and output combined): about 262,144 tokens. Maximum total upload size per request: 512MB (all files and message content combined). Each uploaded file must be smaller than 100 MB; for a larger file, please use the GT Agent desktop software. Accepted image formats: bmp, gif, heic, heif, jpeg, jpg, png, tif, tiff, webp. Maximum number of images: 3600. Maximum size per image: 64MB. Maximum total image size per message: 64MB (text and history share the same request size limit; uploads close to the limit may still be rejected). Image dimensions: at least 200 pixels per side. Maximum image resolution: 36,000,000 pixels in total. Output content formats: Text Pricing: Low Continuous conversation: Not supported AI can make mistakes. Please verify important information.