Google prices Gemini Decisions at 10 cents a million tokens, with no output charge
Bounded agent work, the routing and grading calls every AI system makes thousands of times a day, is being priced toward zero by the two biggest distribution owners in AI, days before Anthropic prices an IPO on premium tokens.
Vincent Jiangtwo-minute read
Google DeepMind opened the Gemini Decisions API on Wednesday, and its price card marks a break from how models have been sold. The endpoint's gemini-decision-1 model scores up to 1,000 yes/no, multiple-choice and rubric questions in a single forward pass over one multimodal input of up to 256K tokens, returning confidence scores in 150 to 250 milliseconds 1.
The pricing is the news: $0.10 per million input tokens, zero output tokens billed, no thinking-token, OCR or transcription charges 1. Against its own cheapest chat tier, Gemini 3.5 Flash-Lite, run through the Interactions API, Google's standard chat-completion endpoint, Google+2.63% — Google, up 2.63 percent today says Decision 1 cost $0.88 on its shared benchmark suite versus $4.14, a 4.7x self-undercut 5. Google is also its own first customer: Gemini Notebook, its research notebook product, uses the endpoint to route queries to the right model tier, at up to 40% lower latency on everyday requests 1.
Everyone sells bounded judgment now
Google is not alone in repricing judgment. Meta−0.87% — Meta, down 0.87 percent today launched Muse Ember on November 9 at the same $0.10 per million input tokens, plus $0.50 per million output 3. OpenAI's$1.18T — OpenAI, private, latest valuation $1.18T Decisions API, generally available since October 22 on GPT-6 Luna, also charges $0.10 input with free output, and offers $0.01 cache reads for reused reference material 5. TypeSafe's Jev, a cheaper classification model, undercuts them at $0.042; Nace AI's Drex Max, a document and tool-use model, sits at $0.03 5.
Five vendors price decision calls at ten cents or less
Five vendors at ten cents or less; three cluster at ten cents
Data
| Value | |
|---|---|
| Nace Drex Max | $0.03 |
| TypeSafe Jev | $0.04 |
| Gemini Decision 1 | $0.1 |
| OpenAI GPT-6 Luna | $0.1 |
| Meta Muse Ember | $0.1 |
Anthropic prices an IPO into a ten-cent market
The timing lands hardest on Anthropic$350B — Anthropic, private, latest valuation $350B, which prices its Nasdaq IPO on November 18: 825 million shares at $70 to $78, up to a $1.405 trillion fully diluted valuation, off $33.24 billion of nine-month revenue 4. That valuation assumes premium token prices hold. But an IPO buyer underwriting $33 billion of revenue on expensive tokens is watching the two biggest distribution owners in AI converge on ten cents for routing, policy checks and grading, the calls every agent makes thousands of times a day 4.
At 10,000 inquiries, model fees come to $2.25
TheStreet's arithmetic shows why: 10,000 customer inquiries at 1,000 input and 250 output tokens each run about $2.25 in Muse Ember fees 3. At that level, model cost stops being the line item; agent plumbing becomes the product.
The number to watch is Google's blended price per token: whether Gemini revenue grows faster than the average price per call falls. Output-free pricing for agent-control work already has three adopters, with Google, OpenAI and TypeSafe all billing $0 output 5. A fourth would make it the standard.
Get Model by email, free
The AI labs, their models, and the business behind them. In our daily email, every weekday.
