Gemma 4 12B (Reasoning)

Gemma 4 12B (Reasoning)

Google · Released Jun 3, 2026 Professional New
Intelligence #10 / 576
82.0 Our Score
Speed #65 / 271
155.1 tokens / sec
Input
— Not priced
Output
— Not priced
Context
— Not reported

Analysis Summary

Gemma 4 12B (Reasoning) is a mid-sized open-weight model from Google with a reasoning-mode configuration. Its instruction-following benchmark is among the stronger results in this tier, and its long-context reasoning score is reasonable, making it a viable option for structured content tasks.

For businesses, it suits content generation, document summarisation, and instruction-heavy workflows where cost efficiency matters more than frontier-level reasoning. Its coding and agentic scores are below average for the field, so it is not well suited to software engineering or autonomous agent pipelines.

No pricing data is available, but as an open-weight model it can be self-hosted, which is a meaningful cost advantage for teams with infrastructure. It is a practical choice for content-focused teams that want a capable, deployable model without API dependency.

Assessed June 9, 2026

Editorial notes

Gemma 4 12B (Reasoning) from Google offers moderate intelligence with strong instruction-following scores, but limited coding and agentic capability for its size.

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?

Performance Profile

Intelligence4.4Technical3.5Value0Content7.3
Intelligence 4.4/10
Technical 3.5/10
Content 7.3/10
Value 0/10

How Gemma 4 12B (Reasoning) compares

Gemma 4 12B (Reasoning) ranks #113 of 378 AI models we track for overall intelligence, #115 of 315 for coding, #149 of 289 for agentic tasks. Gemma 4 12B (Reasoning) is currently free to use via OpenRouter.

Performance Indices

Source: Artificial Analysis

29.2 Intelligence Index
24.9 Coding Index
27.2 Agentic Index

This model was released recently. Independent benchmark evaluations are typically completed within days of release, so these figures are preliminary and are likely to be updated as testing is finalised.

Benchmark Scores

Intelligence

GPQA Diamond 75.3% Graduate-level scientific reasoning
HLE 14.8% Humanity's Last Exam
SciCode 38.2% Scientific computing

Technical

TerminalBench Hard 18.2% Agentic terminal tasks
τ²-Bench 36.3% Conversational agent benchmark

Content

IFBench 73.5% Instruction following
LCR 55.3% Long-context reasoning

Benchmark data from Artificial Analysis and Hugging Face

How does Gemma 4 12B (Reasoning) stack up?

Compare side-by-side with other professional models.

Compare Models

Model Information

ProviderGoogle
Release Date June 3, 2026
Status Active

Leaderboard Categories

Frequently asked questions about Gemma 4 12B (Reasoning)

How much does Gemma 4 12B (Reasoning) cost?

Gemma 4 12B (Reasoning) is currently available for free via OpenRouter.

Is Gemma 4 12B (Reasoning) good for coding?

On our coding benchmark index, Gemma 4 12B (Reasoning) ranks #115 of 315 models, placing it in the broader range of the field for code generation and debugging.

Who created Gemma 4 12B (Reasoning)?

Gemma 4 12B (Reasoning) is developed by Google and was released on June 3, 2026.