The Open-Source AI Stack

RSS

Models · gemini-3

Gemini 3 Pro

Proprietary Google DeepMind · 2025-11-18 · Proprietary

DeepMind's flagship Gemini 3, announced by Sundar Pichai November 18 2025 with broader rollout on November 20. Shipped with a Deep Think mode and an agentic Gemini Agent surface, plus the new Antigravity platform for agent-first development.

Cost

$2.00 / Mtok input

$12.00 / Mtok output

· as of 2026-05-21

Speed

132.9 tok/sec output

29181 ms TTFT

· as of 2026-05-21

Architecture

Schema-generated from data/models.yaml. Every label is auditable against the model's sources.

Specs

Architecture: unknown
Total params: not disclosed
Active params: not disclosed
Context window: 1.0M tokens
Attention: unknown
Position encoding: unknown
Post-training: rlhf
OSI-approved: no
Data released: no
Training code: not released

Benchmarks

Each score carries the date it was published; we never infer or interpolate missing scores.

General reasoning

MMLU-Pro	89.8	as of 2026-05-21	source ↗
GPQA-Diamond	91.9	as of 2025-11-18	source ↗

Code

SWE-Bench Verified	76.2	as of 2025-11-18	source ↗
LiveCodeBench	91.7	as of 2026-05-21	source ↗

Math

AIME 2025

95.7

as of 2026-05-21

Available quantizations

None. The weights are not distributed, so there are no public quantizations.

Notable innovations

· Deep Think reasoning mode
· Antigravity agent-development platform
· Integration across Search AI Mode, Gemini app, Vertex AI on day one

Lineage

Flagship of the Gemini 3 generation; ships with the new Deep Think mode.

Derived from

Gemini 2.5 Pro 2025-03-25

Derivatives

Sources