Google Launches Gemini 3.1 Flash-Lite as Gemini 3 Pro Dies

Google dropped two pieces of Gemini news on the same day: a quiet model launch and a loud deprecation deadline. Gemini 3.1 Flash-Lite Preview appeared in the API changelog on March 3 as "the first Flash-Lite model in the Gemini 3 series." Hours later, Google's Logan Kilpatrick announced that Gemini 3 Pro shuts down on March 9 - giving developers six days to migrate.


New model	Gemini 3.1 Flash-Lite Preview (`gemini-3.1-flash-lite-preview`)
Pricing	$0.25/1M input tokens, $1.50/1M output (free tier available)
Position	Most cost-efficient model in the Gemini 3 lineup
Deprecation	Gemini 3 Pro Preview shuts down March 9; `gemini-pro-latest` alias moves to 3.1 Pro on March 6
Missing	No Gemini 3.1 Flash (text) announced yet - a gap in the lineup

Flash-Lite: Cheap and Fast, Not Frontier

The new Flash-Lite was first spotted on Vertex AI before the changelog went live. Google describes it as "optimized for high-volume agentic tasks, translation, and simple data processing" - the same positioning as previous Flash-Lite models, but now running on Gemini 3 architecture.

At $0.25 per million input tokens, it's a tenth the price of Gemini 3.1 Pro ($2/1M) and comes with batch pricing at 50% discount. The model supports text, image, video, audio, and PDF input with 1M token context, but outputs text only.

The Flash-Lite tier has existed since Gemini 2.0 (January 2025), sitting below Flash in Google's hierarchy: Pro > Flash > Flash-Lite. Each step trades quality for speed and cost. What's notable here is the version jump - there was no "Gemini 3.0 Flash-Lite." Google skipped straight from 2.5 Flash-Lite to 3.1.

Gemini 3 Pro: Six Days to Say Goodbye

The deprecation announcement came from both Logan Kilpatrick and Patrick Loeber on the same day:

March 6: The gemini-pro-latest alias automatically moves to Gemini 3.1 Pro Preview
March 9: Gemini 3 Pro Preview is permanently shut down on AI Studio and the Gemini API
March 23: Vertex AI follows

Kilpatrick was blunt about the reason: "we need to defragment compute so sadly can't keep it around, the frontier presses forward."

Developers in the migration thread are not happy. Multiple users pointed out that Google's own deprecation policy requires "at least two weeks notice" - but the March 3 announcement for a March 9 shutdown gives only six days (or eleven if counted from the original February 26 notice by Loeber). Others complained about being forced onto 3.1 Pro while it still has API stability issues: "It's crazy to deprecate a model that was working, to replace it by an unusable model."

GitHub also confirmed the deprecation, affecting Copilot users who had Gemini 3 Pro as a model option.

The Gap in the Lineup

The March 3 launch creates a curious gap in Google's Gemini 3 model lineup:

Tier	Gemini 3.0	Gemini 3.1
Pro	December 2025 (dying March 9)	February 19, 2026
Flash	December 2025	Not announced
Flash Image	-	February 26, 2026 (Nano Banana 2)
Flash-Lite	Never existed	March 3, 2026

There's no Gemini 3.1 Flash text model. Google launched the Pro, then the image-generation Flash variant (Nano Banana 2), then jumped to Flash-Lite - skipping the middle of its own tier stack. Whether a Gemini 3.1 Flash is coming or whether Flash-Lite is meant to fill that role remains unclear. Google hasn't commented.

For developers, the immediate priority is the March 9 deadline. If you're on Gemini 3 Pro, migrate to 3.1 Pro Preview now - or wait for the alias switch on March 6 and test before the cutoff. If you're building high-volume pipelines, the new Flash-Lite is the cheapest option in the Gemini 3 family.

Sources:

Flash-Lite: Cheap and Fast, Not Frontier

Gemini 3 Pro: Six Days to Say Goodbye

The Gap in the Lineup

Google Analytics