Short answer
Gemini 3 Pro Preview is no longer callable through the Gemini API: Google’s retired-model page identifies its model ID as gemini-3-pro-preview and says it was shut down on March 9, 2026. The page directs users to Gemini 4 Argon to avoid service disruption. Treat Argon as a candidate to evaluate, not a proven drop-in replacement. (Google’s retired-model page)
Google launched Gemini 3 Pro in preview on November 18, 2025. The launch announcement highlighted reasoning, multimodal understanding, coding, tool use, and agentic tasks. Those are historical product claims, not evidence that another model will perform the same way on your workload. (Google’s Gemini 3 launch announcement)
Model name, API ID, and lifecycle
| Name | Identifier or listing | What the cited Google documentation says |
|---|---|---|
| Gemini 3 Pro Preview | Gemini API ID: gemini-3-pro-preview |
The retired-model page says it was shut down March 9, 2026. Do not use this ID for a new Gemini API integration. |
| Gemini 3.1 Pro | Google Cloud model listing: Gemini 3.1 Pro; deprecations table names gemini-3.1-pro-preview as a replacement |
The Cloud listing places Gemini 3.1 Pro under preview models. This is a distinct model name, not another name for Gemini 3 Pro. The cited listing does not establish Gemini API availability or compatibility. |
| Gemini 4 Argon | Gemini API ID: gemini-4-argon |
Google’s current Gemini API guide says Argon is available through the Gemini API and recommends it on the retired Gemini 3 Pro Preview page. It is a candidate to test, not a compatibility guarantee. |
Google’s deprecations table names gemini-3.1-pro-preview as the replacement for gemini-3-pro-preview. Separately, the retired-model page directs users to Gemini 4 Argon. These are the replacement references in those documents; neither establishes that the named models are interchangeable. For a new integration, follow the current API guide and test the target model against your application. (Google deprecations table; retired-model page; Google Cloud model listing)
What Gemini 3 Pro Preview was designed to do
Google positioned Gemini 3 Pro for complex reasoning, multimodal understanding, coding, tool use, and agentic work. The launch announcement described a 1-million-token context window and discussed text, images, video, audio, and code. These are launch-announcement descriptions, not a guarantee that every modality or limit applied identically across API channels. (Google’s Gemini 3 launch announcement)
The retired Gemini API model page listed text, image, video, audio, and PDF inputs, with text output. It also listed function calling and code execution as supported capabilities. These are historical specifications for the retired model, not specifications for a successor. (Retired Gemini 3 Pro Preview API page)
In its November 2025 announcement, Google reported 37.5% on Humanity’s Last Exam without tools and 76.2% on SWE-bench Verified. These are vendor-reported launch figures; the supplied announcement does not include the evaluation methodology. Use them as historical context, not as independent or current comparisons. (Google’s Gemini 3 launch announcement)
What to evaluate for a production API
Start with Gemini 4 Argon as a candidate to test. Google’s Gemini API guide identifies its model ID as gemini-4-argon, says it is available through the Gemini API, and positions it for complex coding, agentic, math, and science tasks. Google’s Cloud model listing places Gemini 4 Argon among generally available Gemini models and Gemini 3.1 Pro among preview models. That status distinction is specific to the Cloud listing; it does not establish identical release terms for the Gemini API. None of these descriptions proves equivalent output or migration compatibility. (Gemini API guide; Google Cloud model listing)
For the Gemini API, Google’s guide lists introductory Argon pricing through November 30, 2026 at $2 per 1 million input tokens and $10 per 1 million output tokens; it lists $4 per 1 million input tokens and $20 per 1 million output tokens thereafter. These are Gemini API prices, not a quote for Vertex AI or a consumer subscription. Check the applicable channel’s current pricing when estimating deployment cost. (Gemini API guide)
Before committing, test representative requests through the API channel you plan to use. Compare task success, structured-output validity, tool and function behavior, latency, token use, and failure handling. Confirm that the channel supports the input types and limits your application needs, and calculate cost using its applicable pricing. Keep the retired gemini-3-pro-preview ID out of new deployments, and do not treat Gemini 3 Pro, Gemini 3.1 Pro, and Gemini 3 Flash as the same model.
Sources
Gemini 3 Pro preview — Gemini API
Gemini deprecations — Gemini API
A new era of intelligence with Gemini 3