Gemma 4 26B A4B API

Google DeepMind, open modelgemma-4-26b-a4bModel source: Hugging Face

Understands text and images at once without being a heavy model.

Gemma 4 26B A4B is Google DeepMind's open mixture-of-experts model: 26 billion parameters with only 3.8 billion active per token, so it understands text and images at once without the cost of a heavy model. On GPU Flow it costs €0.12 per million input tokens and €0.30 per million output.

Reads in one go

≈ 70 pages

32,768 tokens

Brain size

Mid-sized

26B MoE, 3.8B active, NVFP4

Where it runs

Spain (EU)

MIG B200

Understands

Text and images

Reasoning, Tools, Vision

What it is good at

Reasoning and tool calling at low cost. Thinks through a task and calls your functions without the price of a large model.
Text and images in the same request. Send a screenshot, a photo or a chart along with your question and get the answer in text.
Fast answers. Its small active footprint keeps responses quick, within a context of about 70 pages.

It answers in text and does not accept audio.

Pricing

€ per million tokens, VAT included

You send (input)0.12 €

It answers (output)0.30 €

Re-reading repeats (cache)0.012 €

1,000 chat answers~0.15 €

Analysing a 200 page contract~0.01 €

try it with 5 € freeActivated with your card. Enough for about 33,333 answers.

For your technical team

Compatible with the OpenAI SDK: change the base URL and the key, and that is it.

curl https://api.gpuflow.ai/v1/chat/completions \
  -H "Authorization: Bearer $GPUFLOW_API_KEY" \
  -d '{ "model": "gemma-4-26b-a4b",
        "messages": [{ "role": "user", "content": "Hola" }] }'

Context and limits

Context: 32,768 tokens per request, about 70 pages.
Requests per minute: 1,000 per key, with bursts up to double; raisable to 5,000 from the dashboard. 5,000 in total per account.
Tokens per minute: 2,000,000 input (cached tokens do not count) and 400,000 output, per key.
Prompt per request: Up to 256,000 tokens, raisable on request.
If you go over: A 429 error with a Retry-After header telling you how many seconds to wait.

Gemma 4 26B A4B compared with similar models

	Gemma 4 26B A4B	Qwen3.8 27B	Nemotron 3.5 Lightning 30B
Best for	Text and images, reasoning and tools at low cost	Coding, agents, structured output	Fast reasoning and tool use at low cost
Context	≈ 70 pages	≈ 210 pages	≈ 280 pages
Input / output, € per million	0.12 / 0.30	0.10 / 0.30	0.12 / 0.30
1,000 chat answers	~0.15 €	~0.14 €	~0.15 €

Three mid-size models at similar prices: Gemma 4 26B for Google-ecosystem familiarity and images, Qwen3.8 27B for coding, Nemotron 3.5 Lightning for fast reasoning with a long context.

Frequently asked questions

Is Gemma 4 26B A4B free?

The model weights are open source; running them costs GPU time. On GPU Flow you pay €0.12 per million input tokens, and every new account gets €5 of credit when it adds a card.

What does the Gemma 4 26B A4B API cost?

€0.12 per million input tokens, €0.30 per million output and €0.012 per million cached input. VAT included, no monthly fee.

Does Gemma 4 26B A4B support function calling?

Yes, with the standard OpenAI tools parameter.

Can I use Gemma 4 26B A4B with Claude Code or Cursor?

Yes. Claude Code connects through the Anthropic-compatible endpoint (/v1/messages); Cursor and other tools through the OpenAI-compatible one (/v1/chat/completions).

Where does it run?

On our NVIDIA B200 GPUs, in a data centre in Spain.

Does it accept images?

Yes: send them as in the OpenAI API, as an image_url part inside content. Each image counts as input tokens (up to about 280), and everything has to fit in the 32,768-token context. It answers in text and does not accept audio.