Gemma 4 26B A4B
Gemma 4 26B A4B is the mixture-of-experts member of Google's Gemma 4 family, with about 26B total and 3.8B active parameters (8 of 128 experts plus a shared expert). It shares the family's 256K context, image input and thinking mode under Apache 2.0. Its small active size makes generation fast relative to its memory footprint.
Specifications
- Context window
- 256Kabout 197,000 words, or 390 pages
- Max output
- Not published
- Input price
- $0.07per 1M tokens
- Output price
- $0.34per 1M tokens
- Cached input
- Not published
- Weights
- OpenApache-2.0
- Parameters
- 25.8B total, 3.8B activeMoE
- Input types
- Text, Images, Video
- Output types
- Text
- Reasoning mode
- Yes
- Tool calling
- Yes
- Released
- Mar 31, 2026Current
DeepInfra serverless price for gemma-4-26B-A4B-it (256K context).
What it costs in practice
| Workload | Monthly cost at list price |
|---|---|
| Support chatbot30,000 conversations a month, 2,000 input and 400 output tokens each | $8.28 |
| Document summaries2,000 documents a month, 12,000 input and 800 output tokens each | $2.22 |
| Coding agent1,000 tasks a month, 60,000 input and 6,000 output tokens each | $6.24 |
Standard rates, no caching or batch discounts. Run your own numbers.
Strengths
- Only 3.8B parameters active per token
- Apache 2.0 license
- Low hosted prices
Trade-offs to weigh
- All 25B parameters must be loaded despite the small active count
- MoE models can be harder to fine-tune than dense ones
Where it fits
Google positions Gemma 4 26B A4B for:
Where you can use it
- Hugging Face
- Ollama
- DeepInfra
Run it yourself
Released under Apache-2.0. Approximate memory needed just to load the weights:
| Precision | Memory for weights |
|---|---|
| 16-bit (BF16/FP16) | 52 GB |
| 8-bit (FP8/INT8) | 26 GB |
| 4-bit (Q4_K_M, about 4.85 bits) | 16 GB |
This is a mixture-of-experts model: all 25.8B total parameters must sit in memory, but only about 3.8B are used per token, so it runs faster than a dense model of the same size.
Add memory for the KV cache (it grows with context length and concurrent users) and 1-2 GB of runtime overhead. Size it precisely in the VRAM calculator.
Alternatives to consider
| Model | Tier | Context | Input / output per 1M | |
|---|---|---|---|---|
| gpt-oss-20bOpenAI | Compact | 128K | $0.075 / $0.30 | Compare |
| Ministral 3 8BMistral AI | Compact | 256K | $0.15 / $0.15 | Compare |
| Granite 4.2 8BIBM | Compact | 128K | $0.06 / $0.25 | Compare |
| Ministral 3 3BMistral AI | Compact | 256K | $0.10 / $0.10 | Compare |
Frequently asked questions
How much does Gemma 4 26B A4B cost?
Gemma 4 26B A4B lists at $0.07 per million input tokens and $0.34 per million output tokens (checked October 10, 2026). DeepInfra serverless price for gemma-4-26B-A4B-it (256K context). Confirm current rates on the provider's pricing page before you commit.
What is the context window of Gemma 4 26B A4B?
Gemma 4 26B A4B accepts up to 262,144 tokens of context (about 197,000 words, or 390 pages). Long prompts cost more and can be slower, so most apps send only what the task needs.
Is Gemma 4 26B A4B open source?
Gemma 4 26B A4B is an open-weight model released under the Apache-2.0 license. You can download the weights from Hugging Face (google/gemma-4-26B-A4B-it) and run or fine-tune them yourself, within the license terms. Open weights isn't the same as fully open source: training data and code may not be published.
Can I run Gemma 4 26B A4B locally?
Yes, if you have enough memory. The weights alone need about 16 GB at 4-bit quantization and about 52 GB at 16-bit, plus room for the KV cache. Use the VRAM calculator on LLModel.com to check a specific GPU or Mac.
What is Gemma 4 26B A4B best for?
Google positions Gemma 4 26B A4B for low latency, self-hosting, high volume, multimodal. It accepts text, images, video as input, has a reasoning (thinking) mode and supports tool calling. Test it on a sample of your own tasks before you choose it for production.
Sources
- huggingface.co/google/gemma-4-26B-A4B-it
- deepinfra.com/pricing
- ai.google.dev/gemma/docs/releases
- huggingface.co/google/gemma-4-31B-it
- huggingface.co/google/gemma-4-26B-A4B-it/raw/main/config.json
- huggingface.co/api/models/google/gemma-4-26B-A4B-it
- ollama.com/library
Facts checked against these pages on October 10, 2026. Spot an error? Tell us.
Still choosing a model?
Send us your use case and constraints. We'll come back with a shortlist and an evaluation plan you can run on your own data.
- Free and vendor-neutral: no model provider pays for placement.
- Your details go only to partners you agree to.