Gemma 4 26B A4Bon Krater.ai

Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at

Use Gemma 4 26B A4B

Released April 3, 2026

Specifications

Gemma 4 26B A4B at a glance

Context

262.1K tokens

Max output

235.9K tokens

Input

Image, Text, Video

Output

Text

Input price

$0.09 / 1M

Output price

$0.30 / 1M

Krater rank

#76

Image inputTool callingStructured outputsReasoning support

Is Gemini down?

Gemini down for you? See Google incidents and Krater's own Gemini checks, refreshed every 5 minutes.

Live Gemini status

Capabilities

What it is good at

Catalog capabilities

  • Accepts image input
  • Generates text responses
  • Supports tool calling
  • Supports structured outputs

Good for

Image understandingLong documentsUse with toolsStructured responsesDeep analysis

Benchmarks

Catalog data, not Krater benchmarks.

Coding index score

39.3

Agentic index score

2.9

Intelligence index score

16.7

Gemma 4 26B A4B

Chat with Gemma 4 26B A4B in Krater.

Ask anything...

Compare

Compare Gemma 4 26B A4B

Gemma 4 26B A4B vs Gemma 4 31B (batch)

Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output.

Open comparison

Gemma 4 26B A4B vs Gemma 4 31B (free)

Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output.

Open comparison

Gemma 4 26B A4B vs Gemma 4 26B A4B (free)

Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind.

Open comparison

FAQ

Questions about Gemma 4 26B A4B

Clear answers based on the model catalog and Krater access.

Gemma 4 26B A4B supports a context window of 262,144 tokens.

Ready to use Gemma 4 26B A4B?

Access leading models in one focused workspace.

Start with Gemma 4 26B A4B