What GPU do I need to run microsoft/Mage-VL?
4.7B parameters, published in BF16. View on Hugging Face
Set up Mage-VLMage-VL is published by microsoft on Hugging Face, with 117,789 downloads and 391 likes to date. It's a MageVLForConditionalGeneration model built for image-text-to-text, published natively in BF16.
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 8.8 GB | 10.6 GB | RTX 3060 | 1 | $0.110/hr |
| FP8 (quantized) | 4.4 GB | 5.3 GB | RTX 4000 SFF Ada | 1 | $0.198/hr |
| INT4 (quantized) | 2.2 GB | 2.6 GB | RTX 3060 | 1 | $0.110/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run Mage-VL at its published (BF16) precision: 1× RTX 3060, at $0.110/hr per GPU ($0.110/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
Mage-VL: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Grouped-query attention: every layer caches keys and values for a few shared KV heads. Cached per token: 147,456 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.56 GB |
| 32K | 4.5 GB |
| 128K | 18.0 GB |
| 256K (model maximum) | 36.0 GB |
Computed from the layer, head and window counts in the model's own configuration (microsoft/Mage-VL), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
Mage-VL: common questions
Does Mage-VL fit on a 12 GB GPU?
Yes. At BF16 it needs 10.6 GB of VRAM, so a 12 GB card holds it with 1.4 GB to spare. An 8 GB card is not enough for it at BF16.
What is the least VRAM Mage-VL can run in?
2.6 GB, at INT4 (quantized), which fits a 6 GB card, against 10.6 GB at BF16. That row assumes an INT4 checkpoint has actually been published for this model, so check its Hugging Face page before planning around it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More microsoft models
- FrogNano-4B-2609 (4.7B, BF16)
- kosmos-2-patch14-224 (1.7B, F32)
- trocr-base-printed (333M, F32)
- trocr-base-handwritten (333M, F32)
- DialoGPT-small (176M, F16)
Alternatives at this size
Other models for image-text-to-text within about a third of Mage-VL's 4.7B parameters, from other model lines.
- Qwen3.5-4B (4.7B, BF16)
- Unlimited-OCR (3.3B, BF16)
- chandra-ocr-2 (5.3B, BF16)
- DeepSeek-OCR (3.3B, BF16)
- llava-1.5-7b-hf (7.1B, F16)
More on Mage-VL
Fits on a 12 GB GPU at BF16: every model that fits in 12 GB.
Fits on an 8 GB GPU at FP8: every model that fits in 8 GB.
Best vision-language models: how Mage-VL ranks against the rest.
Related reading: H100 pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
