Model reference · open weights
Qari-OCR-VL is an open-weight language model from NAMAA-Space. Qari-OCR-v0.3-VL-2B-Instruct (FP16) weighs 4.4 GB; the smallest configuration that runs it is RTX 3060 12 GB.
For self-hosting, compare this build's requirements below with the GPU you plan to rent.
Summary of the NAMAA-Space/Qari-OCR-v0.3-VL-2B-Instruct model card, 2026-10-01
What it is
| Released by | NAMAA-Space |
|---|---|
| Released | 2025-04-10 |
| Parameters | 2.2B |
| VRAM | 4.4 GB for the weights |
What it runs on
| Card | Requests at once | Context max | Memory | |
|---|---|---|---|---|
| 8K each | 32K each | |||
| RTX 3060 12 GB | 22 | 5 | all 32K | 11.6 GB |
| RTX 4060 Ti 16 GB | 38 | 9 | all 32K | 15.4 GB |
| RTX 3090 24 GB | 72 | 18 | all 32K | 23.4 GB |
| RTX 4090 24 GB | 72 | 18 | all 32K | 23.4 GB |
| RTX 5090 32 GB | 104 | 26 | all 32K | 31.0 GB |
| L40S 48 GB | 159 | 39 | all 32K | 44.0 GB |
| A100 80 GB | 305 | 76 | all 32K | 78.2 GB |
| H100 80 GB | 287 | 71 | all 32K | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | 354 | 88 | all 32K | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | 412 | 103 | all 32K | 107 GB |
| H200 141 GB | 542 | 135 | all 32K | 138 GB |
| B200 180 GB | 703 | 175 | all 32K | 176 GB |
| Requests at once | 8K tokens each | 32K tokens each |
|---|---|---|
| 1 | 6.6 GB | 7.3 GB |
| 5 | 7.6 GB | 11.1 GB |
| 8 | 8.3 GB | 13.9 GB |
| 16 | 10.1 GB | 21.4 GB |
| 32 | 13.9 GB | 36.5 GB |
| 64 | 21.4 GB | 66.5 GB |
One card, with vLLM's small-card settings.