K2-Horizon Guide: VRAM Requirements, GGUF Builds

September 28, 2026

About This Model

K2-Horizon is a family of language models from IFM. What sets it apart from most open-weight models, which publish weights only, is that it aims to be fully open: pre-training and mid-training data, training code, and intermediate checkpoints are all being released. Every size supports a 524,288-token (512K) context, and the family emphasizes reasoning (thinking) mode and agentic use with tool calls.

The family has three models of different sizes:

Model Architecture Parameters active per token Role
K2-Horizon-375B-A23B MoE (375B total) 23B Flagship
K2-Horizon-MoVA-36B-A4B MoE (36B total, with its own “MoVA" attention) 4B Lightweight sparse model
K2-Horizon-32B (Stage 1) Dense 32B 32B Mid-training release (final version to come)

The models this site has covered, with their memory requirements, are listed under “Our Coverage and Data" below.

What Makes It Stand Out

  • The flagship 375B-A23B is positioned as strong on agentic tasks. In the publisher’s comparison it scores 70.2% on Terminal-Bench 2.1 (terminal use), 65.3% on Toolathlon Verified (tool use) and 67.7% on MCPMark (tool use via MCP). The publisher claims it matches or beats open MoE models up to 2.6× its size and competes with closed frontier models.
  • MoVA-36B-A4B scores high for a model that runs only 4B parameters per token. It posts 58.6% on Terminal-Bench 2.1 and 26.8% on tau3-Banking (business tool use), and the publisher says it outscores ~30B dense models and MoE models up to 15× its size. With so little compute per token, it should run faster than a dense model of similar size.
  • Honestly, the 32B (Stage 1) is still a work in progress. Even in the publisher’s own table, it trails similarly sized models from other developers on most benchmarks. It is released as an intermediate checkpoint so researchers can study how capabilities change during training.
  • The NVFP4 builds lose very little. The average across seven evaluations the publisher ran under identical settings:
Model Original (BF16) NVFP4 build Difference
K2-Horizon-375B-A23B 91.7 91.2 −0.5
K2-Horizon-32B (Stage 1) 89.5 88.4 −1.1

Shrinking the weights to roughly 4 bits costs only about one point on average. Note that the NVFP4 builds were evaluated only on non-agentic tasks so far (agentic results are to follow).

Running It Locally

  • The NVFP4 builds require an NVIDIA GPU from the Blackwell generation or later with native NVFP4 support. They will not run on RTX 40-series cards, nor on Hopper-generation data-center GPUs such as the H100 and H200.
  • The 375B-A23B is not realistic for personal hardware. For the original BF16 version, the publisher’s validated setup is a single node with eight H200s. If you want to try the family locally, the GGUF build of MoVA-36B-A4B is the most practical option.
  • The publisher recommends always using reasoning_effort="high" and enabling the dedicated k2_horizon parsers in vLLM and SGLang.
  • The license is Apache-2.0, which allows commercial use.

Sources: model cards for IFM/K2-Horizon-375B-A23B-NVFP4, IFM/K2-Horizon-32B-NVFP4 and IFM/K2-Horizon-MoVA-36B-A4B, as of 2026-09-25. Benchmark figures are as published by IFM.

Our Coverage and Data

Everything Local Model Watch has published about the K2-Horizon family: 3 article(s) covering the base model and its fine-tunes, plus converted builds we tracked after publication. Memory requirements below are computed by this site from file sizes, not quoted from model cards. Part of our model family index.

At a Glance

Item Value
Base model(s) IFM/K2-Horizon-32B-NVFP4, IFM/K2-Horizon-375B-A23B-NVFP4, IFM/K2-Horizon-MoVA-36B-A4B
Publisher IFM (MBZUAI)
License (model card) apache-2.0
Smallest VRAM tier 32GB
Articles 3

Hardware Requirements

Estimated requirements (calculated by Local Model Watch)

Your VRAM Quantization File size Est. memory needed
32GB (RTX 5090, etc.) NVFP4 21.7GB 26.0GB

Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference. What the quantization names mean: glossary.

Figures are for IFM/K2-Horizon-32B-NVFP4. Other sizes in this family are covered in the articles below.

Can You Run It Locally?

Not usable in Ollama, LM Studio and llama.cpp yet.

The publisher ships safetensors only, and llama.cpp’s registry does not list this architecture. llama.cpp would need to add support before these tools can run it. Today it can be run with transformers or vLLM, using the memory figures in the table above.

License — apache-2.0 (Commercial use allowed): Permits commercial use, modification and redistribution. Redistribution requires including the license and stating changes; includes a patent grant.

Compiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found" means we have not seen such a build, not that none exists. License summaries are not legal advice — check the publisher’s original terms before relying on them.

This assessment is for IFM/K2-Horizon-32B-NVFP4.

Articles (the family’s own models first, then newest)

Published Model Type Article
2026-09-23 IFM/K2-Horizon-32B-NVFP4 New Models K2-Horizon-32B-NVFP4 Long-Context Reasoning Model: 32GB+ VRAM
2026-09-23 IFM/K2-Horizon-375B-A23B-NVFP4 New Models K2-Horizon-375B-A23B-NVFP4 Text Generation Model: ~257GB Memory

Repositories

Last updated 2026-09-28 (JST). The explanation at the top of this page was written with the help of AI from the primary sources it cites. The tables and lists under “Our Coverage and Data" are assembled by code from our article log.