2 Apr 2026
gemma-4-E2B-it
“Effective” is a compute label, not the download size
5.1B reported total · text + image + video + audio
Gemma / Dense
gemma-4-E4B-it
More effective capacity, with tables still to place
New to model sizes or GPU memory? Start with weights, parameters and quantization →
E4B is the next Gemma candidate when E2B misses important cases. Its 4.5B effective count comes with about 8B parameters including per-layer embeddings, so the larger model changes both quality and storage.
8Breported total parameters
Denseactive parameters per token
131,072configured token positions
Includes per-layer embeddings; effective compute is smaller. Not a MoE active-parameter count.
Evaluate it for
Multimodal edge evaluation where E2B is too limited and the larger footprint is acceptable.
Choose another path when
Treating the E4B name as an exact four-billion-parameter artifact.
Release context
Why this release matters
Like E2B, it makes conditional embedding storage visible in a small-model setting. This is a useful bridge to the much larger conditional-memory systems introduced later in 2026.
Deployment starting points
Which hardware, how many, at what precision?
Start with an exact artifact and an engine that supports it. A configuration below is evidence of a documented path; its device count does not promise a particular throughput or concurrency.
No certified hardware configuration in this edition
Treating the E4B name as an exact four-billion-parameter artifact.
The pinned model card establishes the architecture. It does not, by itself, prove that a particular GPU count runs this checkpoint at your target context. The capacity screen below helps rule out undersized allocations; a successful load and workload test are still required.
Capacity screen: would the weights fit?
This estimates inference memory for a hypothetical uniform precision. It is useful for rejecting an allocation that is too small. It does not establish a working deployment or estimate training memory.
A uniform-precision GPU count would hide this model’s component layout. Includes per-layer embeddings; effective compute is smaller. Not a MoE active-parameter count. Use the documented artifact and placement path above; no GPU count is inferred here.
Architecture in practice
What the serving engine has to do
Shared tables and request cache have different lifetimes.
Model tables do not multiply with every request.
The lookup and host/device placement are model-specific.
Mechanism schematic based on the pinned model card and configuration. It explains a design principle; it is not a full implementation graph.
Inspect the full checkpoint specifications
| Exact checkpoint | google/gemma-4-E4B-it |
|---|---|
| Text model type | gemma4_text |
| Layers | 42 |
| Hidden width | 2,560 |
| Attention / KV heads | 8 / 2 |
| Head dimension | 256 |
| Routed / selected experts | Not verified / Not verified |
| Configured positions | 131,072 |
| Documented extension | Not recorded for this checkpoint |
| Inputs → output | text, image, video, audio → text |
| License metadata | apache-2.0 |
35 sliding attention layers; 7 full attention layers. These counts describe the configured pattern.
Open weights do not imply unrestricted use. Read the applicable license terms linked from the pinned card.
A useful comparison
What else belongs on the shortlist?
E2B is the smaller placement comparison; 12B takes a different unified multimodal route.
3 Jun 2026
gemma-4-12B-it
One model path for text, vision and audio
11.95B reported total · text + image + video + audio
These are editorial comparison candidates. We have not run a matched quality or serving benchmark, so this is not a ranking.
Prepare to explain it
The interview lesson
Budget the full tables and modality path, then compare memory saved by quantization with quality lost on the actual task.
Attention layouts · Quantization · Courses and worked examples
For a deployment evaluation, record exact weights, precision, engine, device count, interconnect, prompt/output lengths and concurrency. Report task success, errors, TTFT, TPOT and useful throughput together.
Sources and citation
Our assessment is editorial judgment based on the linked architecture and deployment evidence. Model facts are publisher-reported or attributed to runtime maintainers; no independent GPU benchmark was run.
AI Infra Interviews, “gemma-4-E4B-it: release, architecture and deployment”, checked 2026-09-13. Preserve this date and the exact checkpoint when citing.