GOOGLE · SERVERLESS AND DEDICATED · LIVE
gemma-4-26b-a4b-it
Google's efficient Gemma 4 mixture-of-experts with 4B active parameters, strong across languages, streamed at 356 tokens per second.
262K
TOKEN CONTEXT
$0.0525
INPUT, PER MILLION
$0.225
OUTPUT, PER MILLION
Streamed
BYTE FOR BYTE, NO BUFFER
Overview
gemma-4-26b-a4b-it is Google's instruction-tuned Gemma 4 mixture-of-experts: 26B parameters with about 4B active per token. The sparse design is the point. It answers closer to a mid-size dense model while serving at the cost of a small one, with a 262,144-token context window.
The listing tag says multilingual, and that is where teams put it: high-volume support and chat that has to hold quality across languages. At $0.225 per million output tokens it costs less than the median listing for open models of comparable class, which also makes it a common on-ramp before committing traffic to something larger.
Serving on Nyx
Output streams byte for byte as it is generated, 356 tokens per second, with nothing buffering chunks between the engine and your socket. When demand exceeds declared capacity Nyx answers with an immediate 429 and a Retry-After header; requests are never quietly queued into a timeout.
Multilingual chat reuses a lot of prompt scaffolding, and the cross-turn prefix cache removes 31 percent of prefill work on agent traces. For steady traffic, dedicated pods hold guaranteed throughput at a flat hourly price.
OUTPUT PRICE · $/M · LOWER IS BETTER
Independent harness · run weekly · streamed
# OpenAI-compatible: swap the base URL
curl https://api.nyxprovider.com/v1/chat/completions \
-H "Authorization: Bearer $NYX_API_KEY" \
-d '{
"model": "gemma-4-26b-a4b-it",
"stream": true,
"messages": [{"role": "user",
"content": "Reply to this customer in the language they wrote in."}]
}'