Gemma 4 26B A4B — Specifications
Compiled by Mustafa Ihsan from the vendor’s published documentation · Last updated
| Developer | |
|---|---|
| Type | LLM (multimodal) |
| Modality | Text, Image → Text |
| Parameters | 25.2B total / 3.8B active |
| Context window | 256K |
| License | Apache 2.0 |
| Open weights | Yes |
| Released | 2026-03 |
| Input price | $0.15 /1M |
| Output price | $0.60 /1M |
| API providers | Google Cloud (managed API), DeepInfra, OpenRouter (hosted); Gemini API (free tier only); Ollama (local) |
Run it locally
| VRAM (4-bit) | ~14.4 GB |
|---|
Benchmarks
| Google-reported | MMLU Pro 82.6%; GPQA Diamond 82.3%; LiveCodeBench v6 77.1%; AIME 2026 (no tools) 88.3%; MMMU Pro 73.8%. |
|---|
What is Gemma 4 26B A4B?
Gemma 4 26B A4B is the mixture-of-experts model in Google DeepMind’s open Gemma 4 family, released on 31 March 2026 under the Apache 2.0 licence. Its model card lists 25.2 billion parameters in total but only 3.8 billion active per token (8 of 128 experts plus one shared), text and image input and a 256K-token context. Google says it runs almost as fast as a 4B model, and reports 82.6% on MMLU Pro, 82.3% on GPQA Diamond and 77.1% on LiveCodeBench v6, against 85.2%, 84.3% and 80.0% for the dense Gemma 4 31B.
Google Cloud sells it as a managed API at $0.15 per million input tokens and $0.60 output, with cache hits at $0.015; its price page calls it “Gemma 4 26B”. That is the price stored in this entry. DeepInfra lists it for less, at $0.07 and $0.34, and Google’s Gemini API serves gemma-4-26b-a4b-it on its free tier only.
Running it locally takes more memory than the “A4B” suggests. Google notes that all 26 billion parameters must be loaded to keep routing fast, and estimates about 14.4 GB at 4-bit (Q4_0) for the weights alone, before the context window. Ollama’s gemma4:26b-a4b-it-q4_K_M build is an 18 GB download, more than a 16 GB card holds. Size it with the VRAM calculator and see how to run Gemma 4 in Ollama. In the previous generation, the nearest size was Gemma 3 27B.
Gemma 4 26B A4B pricing: API cost per 1M tokens
| Input (per 1M tokens) | $0.150 |
|---|---|
| Output (per 1M tokens) | $0.600 |
| Output/input ratio | 4× |
| Blended (4:1 in:out) | $0.240 per 1M tokens |
What Gemma 4 26B A4B costs per month
Real monthly spend at a 4:1 input-to-output mix — the ratio a typical chat or RAG workload actually produces.
| Workload | Tokens / month | Cost / month |
|---|---|---|
| Side project | 1M in / 0.25M out | $0.30 |
| Small team | 20M in / 5M out | $6.00 |
| Production | 200M in / 50M out | $60 |
Run your own numbers in the AI API cost calculator.
Cheaper alternatives to Gemma 4 26B A4B
| Model | Blended $/1M | You save |
|---|---|---|
| Llama 3.1 8B open | $0.0240 | 90% cheaper |
| Mistral NeMo 12B open | $0.0292 | 88% cheaper |
| Gemma 3 4B open | $0.0600 | 75% cheaper |
Self-host or pay the API?
Gemma 4 26B A4B is open-weight, so you can run it yourself. It needs ~14.4 GB of VRAM at 4-bit. Self-hosting only beats the API once your volume is high enough to keep that hardware busy — the self-hosting vs API calculator works out the break-even point for your token volume.
Frequently asked questions
How much does Gemma 4 26B A4B cost per 1M tokens?
Gemma 4 26B A4B costs $0.150 per 1M input tokens and $0.600 per 1M output tokens. At a typical 4:1 input-to-output mix that blends to about $0.240 per 1M tokens.
How much does Gemma 4 26B A4B cost per month?
A small-team workload of 20M input and 5M output tokens a month costs about $6.00 on Gemma 4 26B A4B. A side project (1M in / 0.25M out) costs roughly $0.30.
What is a cheaper alternative to Gemma 4 26B A4B?
Llama 3.1 8B is the strongest cheaper option in our database at $0.0240 per 1M blended — about 90% less than Gemma 4 26B A4B. It is also open-weight, so self-hosting is an option.
Can I run Gemma 4 26B A4B locally?
Yes. Gemma 4 26B A4B is open-weight and needs about ~14.4 GB of VRAM at 4-bit quantisation.
Why does Gemma 4 26B A4B charge more for output than input?
Output tokens are generated one at a time and cannot be batched the way a prompt can, so they cost the provider more to serve. Gemma 4 26B A4B charges 4× more for output, which is why prompt-heavy workloads are far cheaper to run than generation-heavy ones.
Prices are the published list rates for the model's primary API and are reviewed as providers change them. Volume, batch and cached-input discounts are not included. Compare every model side by side in the AI models database or the LLM leaderboard.