---
format: "aidr-story-markdown/v1"
id: "1e545fc2002f235e2beaf9998d6465fe373a031d72968fff0e1d90be3e58054e"
canonical_url: "https://aidr.today/1e545fc2?lang=en"
title: "Google Gemma Inference Hits 108,000 Tokens per Second via Gemini 4 Argon"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-26T17:38:14.000Z"
category: "Infra"
topics: ["google","inference"]
source_urls: ["https://marketbrief.now/ai/google-gemma-inference-hits-108000-tokens-per-second-via-gemini-4-argon-0be801df","https://huggingnews.com/ai/google-gemma-inference-hits-108000-tokens-per-second-via-gemini-4-argon-0be801df"]
summary: "Google's optimization for the gemma-4-E2B-it model using the vLLM framework and Gemini 4 Argon increased throughput from approximately 37,900 to over 108,000 tokens per second on Nvidia A100 GPUs, a 2.85x improvement. The update was implemented across 8 pull requests and maintained 100% token parity with no drop in quality."
---

# Google Gemma Inference Hits 108,000 Tokens per Second via Gemini 4 Argon

> [Open the canonical story](<https://aidr.today/1e545fc2?lang=en>)

**Published:** 2026-10-26T17:38:14.000Z
**Category:** Infra
**Topics:** google, inference

## Summary

Google's optimization for the gemma\-4\-E2B\-it model using the vLLM framework and Gemini 4 Argon increased throughput from approximately 37,900 to over 108,000 tokens per second on Nvidia A100 GPUs, a 2\.85x improvement\. The update was implemented across 8 pull requests and maintained 100% token parity with no drop in quality\.

## Sources

- [Story source](<https://marketbrief.now/ai/google-gemma-inference-hits-108000-tokens-per-second-via-gemini-4-argon-0be801df>)
- [Story source](<https://huggingnews.com/ai/google-gemma-inference-hits-108000-tokens-per-second-via-gemini-4-argon-0be801df>)