AI Radar

Gemini 3 Pro

Retired · Language, Multimodal, Reasoning · Milestone

Gemini 3 Pro is a multimodal reasoning model in Google's Gemini Pro line and the first Gemini 3 model, released in preview on 18 November 2025 in the Gemini app, AI Mode in Search, Google AI Studio, Vertex AI and Google Antigravity. The model card describes a sparse mixture-of-experts transformer; the API preview was shut down on 9 March 2026. [1] [2]Model Information: Description [3]March 9, 2026 Primary source

Timeline of Gemini 3 Pro →

Claims and evidence

  • Released Released in preview (gemini-3-pro-preview) and made available the same day in the Gemini app and other Google products; never generally available in the Gemini API. Primary source[1] [3]November 18, 2025 [4]Gemini 3 models: preview models
  • Deprecated Date of the deprecation announcement in the Gemini API release notes. Primary source[3]February 26, 2026
  • Retired Primary source[3]March 9, 2026 [4]Gemini 3 models: preview models [5]warning banner
  • Status RetiredRetired in the Gemini API; the gemini-3-pro-preview ID now points to gemini-3.1-pro-preview. Primary source[4]Gemini 3 models: preview models [3]March 9, 2026 [5]warning banner
  • Replaced by Gemini 3.1 ProRecommended replacement as listed on the deprecations page when checked on 2026-10-01. Primary source[4]Gemini 3 models: preview models [3]March 9, 2026 [5]warning banner
  • Successor of Gemini 2.5 Pro Primary source[1]A note from our CEO; Build anything
  • Change · Reasoning Adds an optional Deep Think mode, meant to improve complex problem solving at inference time.Compared with Gemini 2.5 Pro Primary source[2]Model Information: Description [1]
  • Change · Availability Available in AI Mode in Google Search on launch day, which Google says was a first for a Gemini model.Compared with Gemini 2.5 Pro Primary source[1]
  • Change · Other New API behaviour for media resolution, thought signatures and thinking levels.Compared with Gemini 2.5 Pro Primary source[3]November 18, 2025
  • Input text, image, audio, videoThe Gemini API also accepted PDF files. Primary source[5]Supported data types [2]Model Information: Inputs
  • Output text Primary source[5]Supported data types [2]Model Information: Outputs
  • Feature Reasoning modeThinking with thinking levels in the API; optional Deep Think mode announced with the model. Primary source[2]Model Information: Description [5]Capabilities: Thinking [3]November 18, 2025
  • Feature Function calling Primary source[5]Capabilities
  • Feature Tool useCode execution, search grounding and URL context in the Gemini API. Primary source[5]Capabilities
  • Context window 1,048,576 tokensThe model card says up to 1M tokens. Primary source[5]Token limits: input token limit
  • Parameters Not disclosedNot disclosed by Google. No claim madeNo source
  • Access API, consumer appGemini app, AI Mode in Search, Google AI Studio and Gemini API, Vertex AI and Google Antigravity. Primary source[1]A note from our CEO [2]Distribution

Lineage

Predecessors

Successors

Based on

Not derived from another model.

Variants and derived

Siblings

None recorded.

All ancestors

All descendants

Variants

No variants recorded in this record.

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from Google on AI Radar →