3 models · Multimodal (any-to-any) · Transformers.js
Multimodal AI in the browser
3 models handle multimodal in the browser today, from 1.81 GB (Janus-Pro-1B) to 4.07 GB (Gemma-4-E4B-it), a 3.15 GB download at the midpoint. Every one runs with Transformers.js over WebGPU or WebAssembly: no server, no install, no account.
- Models
- 3
- Smallest
- 1.81 GB
- Median
- 3.15 GB
- Largest
- 4.07 GB
Every multimodal model, ranked by WebGPU size
| Model | WebGPU download | WASM download |
|---|---|---|
| Janus-Pro-1B 2.08B params | 1.81 GB q4f16 | 2.11 GB uint8 |
| Gemma-4-E2B-it 5.12B params | 3.15 GB q4f16 | 2.89 GB q8 |
| Gemma-4-E4B-it 8.00B params | 4.07 GB q4f16 | 3.18 GB q8 |
Sizes measured from the HuggingFace API file tree for each model's repo, not estimated. Sorted smallest to largest by the WebGPU headline pick.
Size ladder
The WebGPU download spans 2.2x here: 1.81 GB (Janus-Pro-1B) to 4.07 GB (Gemma-4-E4B-it).
2 of 3 models (Gemma-4-E2B-it, Gemma-4-E4B-it) download a smaller build over WebAssembly than WebGPU: the CPU-fallback quant compresses tighter than the GPU pick there.
None of the 3 models here fit under 100 MB over WebGPU.
Bytes-per-million-params ranges 1.7x within this group: 0.52 MB/M (Gemma-4-E4B-it) to 0.89 MB/M (Janus-Pro-1B).
Notable multimodal models
Other tasks
Or see every browser model grouped by task, or the main GGUF catalog for native runtimes outside the browser.
FAQ
Do all multimodal models here use the same WebGPU quant?
Yes: all 3 models here use the q4f16 WebGPU build.
What would it cost to download every multimodal model here?
9.02 GB total over WebGPU across all 3 models, if you tried every one back to back. Most projects only need the single model that fits the job, not the whole set.
Sizes measured 2026-08-01 from the HuggingFace API. Last validated 2026-09-14. See all browser models or the methodology.