IFM models

3 IFM models from IFM on Hugging Face, from 1.1B to 37.4B parameters, published by IFM in BF16, with contexts from 128K to 512K tokens. The smallest official model, K2-Type-0.9B, needs about 2.4 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 IFM 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
K2-Type-0.9B1.1B2.4 GB1.2 GB0.6 GBRTX 3070$0.088/hr1.8 GB
PAN229.2B65.3 GB32.6 GB16.3 GBA100$1.18/hrnot published for this architecture
K2-Horizon-MoVA-36B-A4B37.4B83.7 GB41.8 GB20.9 GBRTX PRO 6000$1.38/hr6.0 GB

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 (3)

ModelParametersPublished asVRAM neededLive GPU fitEst. $/hr
K2-Type-0.9B1.1BBF162.4 GBRTX 3070$0.088/hr
PAN229.2BBF1665.3 GBA100$1.18/hr
K2-Horizon-MoVA-36B-A4B37.4BBF1683.7 GBRTX PRO 6000$1.38/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.