Ovis models

6 Ovis models from ATH-MaaS on Hugging Face, from 853M to 10.2B parameters, published by ATH-MaaS in BF16, with contexts from 8K to 256K tokens. The smallest official model, OvisOCR2, needs about 1.9 GB of VRAM at its published precision; the cheapest live fit is RTX 3070 at $0.088/hr.

Pick a size

One row per official Ovis size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.

ModelParametersNative VRAMFP8 VRAMINT4 VRAMLive GPU fit (native)Est. $/hrKV cache at 32K
OvisOCR2853M1.9 GB1.0 GB0.5 GBRTX 3070$0.088/hr0.38 GB
Ovis2-1B1.3B2.8 GB1.4 GB0.7 GBRTX 3070$0.088/hrnot published for this architecture
Ovis1.6-Llama3.2-3B4.1B9.3 GB4.6 GB2.3 GBRTX 3060$0.110/hrnot published for this architecture
Ovis2-4B4.6B10.3 GB5.2 GB2.6 GBRTX 3060$0.110/hrnot published for this architecture
Ovis2.5-9B9.2B20.5 GB10.3 GB5.1 GBRTX A5000$0.176/hrnot published for this architecture
Ovis1.6-Gemma2-9B10.2B22.8 GB11.4 GB5.7 GBRTX A5000$0.176/hrnot published for this architecture

VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.

Official models (6)

ModelParametersPublished asVRAM neededLive GPU fitEst. $/hr
OvisOCR2853MBF161.9 GBRTX 3070$0.088/hr
Ovis2-1B1.3BBF162.8 GBRTX 3070$0.088/hr
Ovis1.6-Llama3.2-3B4.1BBF169.3 GBRTX 3060$0.110/hr
Ovis2-4B4.6BBF1610.3 GBRTX 3060$0.110/hr
Ovis2.5-9B9.2BBF1620.5 GBRTX A5000$0.176/hr
Ovis1.6-Gemma2-9B10.2BBF1622.8 GBRTX A5000$0.176/hr

VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. See the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.

Start ComfyUI on a rented GPU.

© 2026 Aquanode. All rights reserved.

All trademarks, logos and brand names are the property of their respective owners.