Gemini 3.1 Pro

Gemini 3.1 Pro is Google DeepMind's flagship reasoning and multimodal model, and evaluating it means looking at two genuinely separate products: the consumer subscription experience inside the Gemini app, and the developer-facing API used to build on top of the model directly. Conflating the two leads to confusing price comparisons, since they're priced on entirely different bases (a flat monthly subscription versus per-token usage).

On the consumer side, Gemini 3.1 Pro is reachable for free at gemini.google.com and in the mobile apps: the free tier includes Gemini 3.5 Flash for everyday use plus a daily allotment of Gemini 3.1 Pro for harder reasoning tasks, along with image generation and a limited number of Deep Research reports per month. Google AI Plus, at $7.99/month, and Google AI Pro, at $19.99/month (the tier that lines up with what ChatGPT Plus and Claude Pro charge), raise those daily limits and unlock heavier features. Google AI Ultra starts at $99.99/month for users who need the highest usage caps and earliest access to new capabilities, with a second, higher Ultra tier above that for the most demanding workloads. For most individual users comparing chat assistants side by side, the AI Pro tier is the most direct equivalent to a ChatGPT Plus or Claude Pro subscription.

On the developer side, the same model is available through the Gemini API and Vertex AI at $2 per million input tokens and $12 per million output tokens, a separate, usage-based price that has nothing to do with the consumer subscription tiers above. A team building a product on Gemini 3.1 Pro pays by token volume, not by seat, and that API price is what actually determines cost at scale for an application rather than the consumer app's monthly fee. The gap between the two pricing models is worth being explicit about, since a headline "$2 per million tokens" figure means nothing to someone just deciding whether to pay for the consumer app, and a "$19.99/month" figure means nothing to a developer estimating API spend at production volume.

Model-wise, Gemini 3.1 Pro is built on a Transformer-based Mixture-of-Experts architecture with a 1 million token context window, which supports use cases like analyzing an entire codebase or a several-hundred-page document in a single prompt without pre-chunking it. Reported benchmark results show it leading on a majority of tracked evaluations, including strong reasoning and coding scores, with native multimodal support across text, image, audio, and video inputs.

The model was still described as in preview rather than generally available as of this review, and reported latency (particularly time-to-first-token on longer reasoning tasks) trails some competitors, along with weaker performance on computer-use-style benchmarks compared to models built specifically around that capability.

For anyone deciding between Gemini and a competing assistant, the practical question splits along the same consumer-versus-API line: someone comparing monthly chat-assistant subscriptions should weigh AI Pro's $19.99/month against ChatGPT Plus or Claude Pro at the same tier, while a team building an application should compare Gemini's per-token API pricing directly against the equivalent API tier of whichever model it's being weighed against, since those two decisions rarely land on the same answer.