Can PaddleOCR VL 1.6 run on the RX 7600?
Yes. At 8K context the best fit is F16, using about 2.5 GB of the 7.6 GB usable on the RX 7600. Estimated speed ceiling: ~150 tokens/sec.
Newer compressed-attention design: long-context figures are an upper bound.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | 1.0 GB Fits | 1.1 GB Fits | 1.5 GB Fits | 3.2 GB Fits | 801 |
| 3-bit (Q3_K_M) | 1.1 GB Fits | 1.2 GB Fits | 1.6 GB Fits | 3.3 GB Fits | 615 |
| 4-bit (Q4_K_M) | 1.2 GB Fits | 1.3 GB Fits | 1.7 GB Fits | 3.4 GB Fits | 492 |
| 5-bit (Q5_K_M) | 1.3 GB Fits | 1.4 GB Fits | 1.8 GB Fits | 3.5 GB Fits | 422 |
| 6-bit (Q6_K) | 1.4 GB Fits | 1.5 GB Fits | 1.9 GB Fits | 3.6 GB Fits | 366 |
| 8-bit (Q8_0) | 1.6 GB Fits | 1.7 GB Fits | 2.1 GB Fits | 3.8 GB Fits | 283 |
| 16-bit (F16/BF16) | 2.5 GB Fits | 2.5 GB Fits | 2.9 GB Fits | 4.6 GB Fits | 150 |
• real GGUF file size. Speed is a bandwidth ceiling (288 GB/s), not a benchmark. Methodology.
More
PaddleOCR VL 1.6 on every GPU and Mac · Every model that runs on the RX 7600 · Other models that fit 8 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (PaddlePaddle/PaddleOCR-VL-1.6), AMD specs. Data as of 2026-11-17.