---
title: "Google: Gemma 4 26B A4B  — price, capability and what beats it · Undominated.ai"
canonical: https://undominated.ai/models/google__gemma-4-26b-a4b-it/
description: "Google: Gemma 4 26B A4B : $0.070/M in, $0.340/M out. Independent capability scores, context-tier pricing, and the models that are both better and cheaper."
---

# Google: Gemma 4 26B A4B  — price, capability and what beats it · Undominated.ai

> Google: Gemma 4 26B A4B : $0.070/M in, $0.340/M out. Independent capability scores, context-tier pricing, and the models that are both better and cheaper.

[Leaderboard](/) / Google

# Gemma 4 26B A4B

Google · released 2026-04-03

 $0.138 per million tokens, balanced Balanced Summarise Chat Code gen Agentic

## DeepSeek V4 Flash 0731 scores higher and costs less — but you would give something up.

**+25.7** on the capability index and **24% cheaper** ($0.105/M against $0.138/M). What you lose:

 - no image, video input

Capability scores do not measure context length, output ceiling or which inputs a model accepts, so a higher score does not mean a drop-in replacement.

The same model via **free** is 100% cheaper (free/M) — same weights, different latency.

### Every price dimension

| Input | $0.070 /M |
| --- | --- |
| Output | $0.340 /M |

### Independent scores

| Intelligence | 26.1 |
| --- | --- |
| Coding | 39.3 |
| Agentic | 11 |
| LMArena Elo | 1434.6 default |

Sources are listed separately rather than averaged. Across the 62 models both have scored they correlate at r = 0.806 — close agreement overall, but four models rank very differently between them, and a blended score would hide exactly those.

### Capability

| Context window | 262K |
| --- | --- |
| Max output | 16K |
| Input modes | image, text, video |
| Tool use | yes |
| Reasoning | optional |

### Provenance

| Price source | openrouter.ai |
| --- | --- |
| Fetched | 2026-08-24 |
| Quality data | verified |

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...