Back to Gap Map Model components / Fine-tuned / chat models

Gemini Pro

closed / Overall score: 5.0(leading)

Gemini Pro is Google DeepMind's higher-capability Gemini tier, the counterpart to Gemini Flash, taking text, image, video, audio and PDF input with a 1M-token input window and a thinking mode. The name is applied to whichever generation currently ships rather than to one fixed release. Weights are closed, and the tier is offered through the Gemini API, Google AI Studio and the Gemini app.

Google still labels the current Pro, Gemini 3.1 Pro, as Preview, and the scores read that checkpoint.

Openness

1 high confidence

1.0

weights: closed
data: closed
code: closed
license: Proprietary(API-only)

Google files Gemini under API and app access and lists Gemma separately as its open models, so the vendor draws the closed line itself. The Gemini API terms bar extracting or replicating the underlying models, including parameter weights, and the current Pro release, Gemini 3.1 Pro, is offered the same way.

https://ai.google.dev/gemini-api/terms recorded 2026-09-29
Gemini API Additional Terms state "You also may not attempt to reverse engineer, extract or replicate any component of the Services", including "parameter weights".
https://deepmind.google/models/gemini/ recorded 2026-09-29
DeepMind's Gemini page lists "Gemini 3.1 Pro" with the line "Best for complex tasks and bringing creative concepts to life" and "3.5 Pro coming soon", and files Gemma separately under "Open models"; no weight download is offered for Gemini.

Adoption

5 medium confidence

5.0

Gemini Pro is offered through the Gemini API, Google AI Studio, the Gemini app and Google's enterprise agent platform. Google publishes no per-model usage or user figure for Pro, so this reading is a judgment from the surfaces it powers rather than a measured count.

https://deepmind.google/models/gemini/ recorded 2026-08-13
Pro tier offered across the Gemini app, Google AI Studio, the Gemini API and Vertex AI; no per-model usage or user figure published

Capability

5 medium confidence

5.0

DeepMind's benchmark table shows Gemini 3.1 Pro ahead of Sonnet 4.6, Opus 4.6 and GPT-5.2 on Humanity's Last Exam, GPQA Diamond, ARC-AGI-2, Terminal-Bench and LiveCodeBench Pro, and behind Opus 4.6 on SWE-bench Verified (80.6% versus 80.8%). Those comparison models have since been superseded, and no independent LMArena placement is available.

https://deepmind.google/models/gemini/ recorded 2026-08-13
Pro presented as the flagship reasoning tier of the Gemini family; a product listing only, carrying no benchmark figures.
https://deepmind.google/models/gemini/pro/ recorded 2026-08-14
DeepMind's Gemini Pro page, Performance table in body text. Gemini 3.1 Pro (Thinking High) - Humanity's Last Exam 44.4% no tools, GPQA Diamond 94.3% no tools, ARC-AGI-2 77.1%, SWE-Bench Verified 80.6% single attempt, LiveCodeBench Pro 2887 Elo, Terminal-Bench 2.0 68.5%, MMMLU 92.6%. Comparison columns Sonnet 4.6, Opus 4.6, GPT-5.2 - Opus 4.6 leads SWE-Bench Verified at 80.8%.

Verified 2026-09-29, except some scores, still awaiting re-confirmation