Gemma 4 26B-A4B IT is a Mixture-of-Experts model from Google's open-weight Gemma 4 family with 26B total but only ~4B active parameters, giving high-throughput inference for advanced reasoning tasks. It is a multimodal, hybrid-thinking model supporting 140+ languages and up to 256K context.
Cost rate
0.1x
Context
262K
Released
Apr 3, 2026
Input
ImageTextVideo
Output
Text
Support
ReasoningTool callingStructured outputs
Performance
Median latency and throughput measured across recent requests.
Throughput74 tok/s
Latency1.71s
Best at
The categories where this model ranks highest.
Mathematical
#52 · top 13%
Image understanding
#63 · top 38%
Legal & Government
#77 · top 20%
Business, Management, & Finance
#93 · top 22%
| Domain | Score | Rank |
|---|---|---|
| CodingCode Arena Elo | 1358 | #126 |
| Software & IT Services | 1476 | #104 |
| Writing, Literature, & Language | 1406 | #117 |
| Life, Physical, & Social Science | 1454 | #107 |
| Entertainment, Sports, & Media | 1407 | #102 |
| Business, Management, & Finance | 1444 | #93 |
| Mathematical | 1480 | #52 |
| Legal & Government | 1464 | #77 |
| Medicine & Healthcare | 1436 | #140 |
| Multimodal | ||
| Image understanding | 1242 | #63 |
Latest news about Gemma 4 26B A4B
Recent articles and community posts, linked to their original sources.


Why Are Programmers Talking About Google Gemini 4 NeonCodex Community2,175
Google Gemini Is a Whole New App Now 🤯 (Full Tutorial)Your AI Workflow4,734
Gemini 4 Neon Is INSANE – Is THIS Google's BEST Model Yet?Bijan Bowen74,279
Gemini Neon Is Here (Personally Tested)Paul J Lipsky168,146
Gemini 4 NeonHacker News1,042774
Google Gemini's "Tasks" Personal Assistant Is Now FreeHacker News5776
Gemini 4 Hydrogen and OrchestratorHacker News509165
Gemini 4 HydrogenHacker News212