Unlimited-OCR-AWQ
AWQ 4-bit (W4A16) quantization of baidu/Unlimited-OCR, a 3B vision-language OCR model that pushes DeepSeek-OCR one step further (one-shot, long-horizon document parsing). This repo quantizes the DeepSeek-V2 MoE text decoder with activation-aware scaling (AWQ) while keeping the vision tower in BF16, so it stays a drop-in transformers model.
Params
3.4 B
Context
32,768
Downloads 30d
1.3 M
Likes
2
Download history
daily snapshots · 34 days
▲ 364 K in the last 30 days (38.1%)
1.4 M1.0 M
Aug 23Sep 2Sep 12Sep 21
1.4 M817 K
Aug 19Aug 30Sep 10Sep 21
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| model.safetensors | i32 | 2.8 GB | 4.1 GB | ✅ Runs comfortably |
| model.safetensors (bf16, full) | bf16 + 32K ctx | 2.8 GB | 5.6 GB | ✅ Runs comfortably |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Specifications
- Architecture
- UnlimitedOCRForCausalLM
- Parameters
- 3.4 B
- Tensor type
- I32
- Context length
- 32,768
- Vocabulary
- 129,280
- Layers / heads
- 12 / 10
- Licence
- mit
- First seen on the Hub
- 2026-06-23
- Base model
- Unlimited-OCR
- Training datasets
- undisclosed
- Added to our catalog
- 2026-08-19
Family
Base model and the most-downloaded derivatives in the catalog.
Compare with any image-text-to-text model