Can LightOnOCR 3 0.8B run on the RX 9070?
Yes. At 8K context the best fit is F16, using about 2.4 GB of the 15.2 GB usable on the RX 9070. Estimated speed ceiling: ~422 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | 1.2 GB Fits | 1.3 GB Fits | 1.5 GB Fits | 2.7 GB Fits | 2001 |
| 3-bit (Q3_K_M) • | 1.4 GB Fits | 1.4 GB Fits | 1.7 GB Fits | 2.8 GB Fits | 1373 |
| 4-bit (Q4_K_M) • | 1.4 GB Fits | 1.5 GB Fits | 1.7 GB Fits | 2.9 GB Fits | 1209 |
| 5-bit (Q5_K_M) • | 1.5 GB Fits | 1.5 GB Fits | 1.8 GB Fits | 2.9 GB Fits | 1107 |
| 6-bit (Q6_K) • | 1.5 GB Fits | 1.6 GB Fits | 1.8 GB Fits | 3.0 GB Fits | 1016 |
| 8-bit (Q8_0) • | 1.7 GB Fits | 1.7 GB Fits | 2.0 GB Fits | 3.1 GB Fits | 788 |
| 16-bit (F16/BF16) • | 2.3 GB Fits | 2.4 GB Fits | 2.7 GB Fits | 3.8 GB Fits | 422 |
• real GGUF file size from prithivMLmods/LightOnOCR-3-0.8B-GGUF. Speed is a bandwidth ceiling (640 GB/s), not a benchmark. Methodology.
More
LightOnOCR 3 0.8B on every GPU and Mac · Every model that runs on the RX 9070 · Other models that fit 16 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (lightonai/LightOnOCR-3-0.8B), AMD specs. Data as of 2026-11-17.