gemini-3-pro

replicate · Chat model

gemini-3-pro is listed here as a chat model from replicate. This page shows simple API pricing, token limits, and capability flags so you can compare it with similar options.

Provider and model identifiers are kept in their original form for accuracy.

replicate-replicate-google-gemini-3-pro

Input

$2.0000 / 1M tokens

Output

$12.0000 / 1M tokens

Cached input

N/A

Context window

N/A

Catalog generated: Jul 13, 2026

Compare this model Open a side-by-side table using this route and a close match from the current catalog. Estimate cost Use token counts, request volume, and cache assumptions to check a realistic monthly bill.

Quick read

Best for

Use this page when you need a fast view of cost, context size, and supported features before testing the model in your own workload.

Things to verify

Always check the provider page for discounts, cache pricing, region rules, and any model limits that may not appear in public metadata.

Pricing

Item	Price
Input	$2.0000 / 1M tokens
Output	$12.0000 / 1M tokens
Embedding	$2.0000 / 1M tokens

Limits

No matching data

Explore benchmark family Review source-backed benchmark rows for the attached base model family. Check data sources See how pricing data is generated and what still needs provider verification. API vs subscription Learn when API pricing matters compared with ChatGPT, Claude Code, Cursor, or Copilot subscriptions. Cache and batch pricing Understand cached input, cache write, batch, priority, and flex pricing rows shown on this page.

Capabilities

Capability	Supported
Vision	Supported
Function calling	Supported
Parallel function calling	Supported
Tool choice	Supported
Prompt caching	-
Reasoning	-
Response schema	Supported
System messages	Supported
Audio input	-
Audio output	-
Web search	-
PDF input	-
Video input	-

Benchmarks

Most benchmark rows are attached to the base model family rather than this provider route. Open benchmark explorer

Benchmark	Score	Metric	Scope	Checked	Source
Humanity's Last Exam	37.5% (no tools)	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
ARC-AGI-2	31.1%	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
GPQA Diamond	91.9%	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
Terminal-Bench 2.0	56.9%	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
SWE-bench Verified	76.2% (single attempt)	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
LiveCodeBench Pro	2439 Elo	Elo	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
MMMU-Pro	81.0%	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
MRCR v2	77.0% (128k average)	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
MRCR v2	26.3% (1M pointwise)	accuracy	Base model: Gemini 3 Pro (Gemini 3 Pro Thinking (High))	2026-05-31	Link
SWE-bench Verified	69.60%	% resolved	Base model: Gemini 3 Pro (Gemini 3 Pro)	2026-05-31	Link
LMArena Text Arena (English)	1489±5	Arena Elo	Base model: Gemini 3 Pro (gemini-3-pro)	2026-05-31	Link
MMLU-Pro	89.8%	accuracy	Base model: Gemini 3 Pro Preview (Gemini 3 Pro Preview (high))	2026-05-31	Link
MMLU-Pro	89.5%	accuracy	Base model: Gemini 3 Pro Preview (Gemini 3 Pro Preview (low))	2026-05-31	Link

Sources

Source links	LiteLLM repository LiteLLM model cost map
Pricing data	LiteLLM model cost map
Synced at	2026-05-28
Catalog generated	2026-07-13T14:21:55.110Z