Google
Gemma 4 26B A4B
前沿
Apr 2026发布日期256K tokens上下文Apache-2.0许可证82 优秀质量
Gemma 4 26B A4B (25.200000762939453B parameters) requires approximately 20.8 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 3.799999952316284B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 24 GB of VRAM.
快速开始
— 复制粘贴即可本地运行Copy-paste commands to run Gemma 4 26B A4B on your machine.
Run
ollama run gemma4:26bQuick specs
Parameters25.2B (3.8B active)
Architecturemoe (MoE)
Context256K tokens
Modalitytext
Min RAM9.8 GB
Rec. RAM15.4 GB (Q4_K_M)
LicenseApache-2.0
FamilyGemma
✓ Code✓ Chat✓ Reasoning
About this model
- •MoE: 128 experts, 8 active per token
- •256K context window
- •#3 open model on Arena
- •Apache 2.0 license
- •89% AIME 2026
相关模型
快速推荐
最佳硬件
Gemma 4 26B A4B 的最佳选择
运行此模型
量化选项
各量化级别的 VRAM 估算
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 9.8 GB | Low | — |
Q3_K_S | 3 | 12.3 GB | Low | — |
NVFP4 | 4 | 14.1 GB | Medium | — |
Q4_K_M | 4 | 15.4 GB | Medium | — |
Q5_K_M | 5 | 18.1 GB | High | — |
Q6_K | 6 | 20.7 GB | High | — |
Q8_0 | 8 | 27.0 GB | Very High | — |
F16 | 16 | 51.7 GB | Maximum | — |
Quality benchmarks
Gemma 4 26B A4B benchmark scores
Coding
SWE-bench Verified—
HumanEval+—
Aider Polyglot—
LiveCodeBench77.1%
Reasoning
MMLU-Pro82.6%
GPQA Diamond82.3%
MATH-500—
ARC Challenge—
Source: official · 2026-04-02
硬件兼容性
全部硬件的适配估算
Computing compatibility...
内存详细分析
Reference: RTX 2060 6GB
Weights15.4 GB
KV Cache3.7 GB
Runtime1.2 GB
Headroom0.6 GB
常见问题
FAQ — Gemma 4 26B A4B
另请参阅