Gemma 4 26B A4B IT
Live·Googlemodel·Served by Google
Open Gemma instruction model for efficient chat and self-hosted deployments
Providers
Every request goes to the cheapest healthy route and fails over to the next.
Loading the routes…
Latency and throughput appear after the first hourly test request to each provider, or once this model has traffic here. Every request goes to the cheapest healthy route and fails over to the next.
Pricing
What a call costs through Rapid Router, at the route requests go to.
What it supports
- Tool calling (function calling)
- JSON mode and structured output
- Image input (vision)
- Audio input
- Extended reasoning
- Streaming responses
- OpenAI SDK, Anthropic SDK and any OpenAI-compatible client
Make a request
The same request you send to OpenAI today, pointed at https://rapidrouter.ai/api/v1.
curl https://rapidrouter.ai/api/v1/chat/completions \
-H "Authorization: Bearer $RAPIDROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "google/gemma-4-26b-a4b-it",
"messages": [{ "role": "user", "content": "Hello" }]
}'Frequently asked
How do I call Gemma 4 26B A4B IT through an OpenAI-compatible API?
Point any OpenAI-compatible client at https://rapidrouter.ai/api/v1 with a Rapid Router API key and set the model to google/gemma-4-26b-a4b-it. The chat completions endpoint is https://rapidrouter.ai/api/v1/chat/completions.
What is the context window of Gemma 4 26B A4B IT?
Gemma 4 26B A4B IT has a context window of 262,144 tokens on Rapid Router, and returns up to 235,929 output tokens per reply.
What can Gemma 4 26B A4B IT do?
On Rapid Router, Gemma 4 26B A4B IT supports tool calling, JSON mode, image input, extended reasoning.
Can I pay for Gemma 4 26B A4B IT in rupees with UPI?
Yes. Rapid Router bills in Indian rupees: top up with UPI or a card, get a GST invoice on every recharge, and Gemma 4 26B A4B IT is debited from that balance at the provider's list price per million tokens.
More Gemini models
Compare and rank
What a month of your own usage costs on it: cost calculator · every head-to-head: comparisons