Aiqre

Model library / IFM

K2-Horizon-MoVA-36B-A4B

K2-Horizon-MoVA-36B-A4B is a sparse Mixture-of-Experts language model with MoVA attention, released by IFM.

37.4B

Mixture of experts, count not published

524,288

Context length in tokens, as published

70 GiB

Weights at the released precision

apache-2.0

Licence declared on the repository

Overview

What this model is.

The model is a Mixture-of-Experts architecture employing Mixture-of-Values attention, with 36 billion parameters and an active 4 billion-parameter compute budget per token. IFM describes it as targeting agentic and reasoning benchmarks, offering frontier-class performance relative to dense and other MoE models.

In practice the model weighs 37.4 billion parameters, stored in bfloat16 format, and supports a native context window of 524,288 tokens. It is released under the Apache-2.0 licence. The repository notes the presence of experts, though the exact count is not disclosed.

Specification

The published shape.

RepositoryIFM/K2-Horizon-MoVA-36B-A4B
PublisherIFM
Published2 September 2026
ArchitectureK2HorizonForCausalLM
Model typek2_horizon
Parameters37,444,792,020
Layers48
Hidden size2,560
Attention heads32, 8 key/value heads
Context length524,288 tokens
Vocabulary250,624 tokens
Weight formatbfloat16
Licenceapache-2.0

Memory

How much VRAM the weights need.

Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.

PrecisionWeightsNotes
bf1669.7 GiBFull precision as released
fp834.9 GiB8-bit, near-lossless on most models
int417.4 GiB4-bit, smallest footprint

The cache costs 192 KiB per token at 16-bit, so the full 524,288-token context of one request needs about 96.0 GiB. Concurrency multiplies that number, not the weights.

Hardware

Which card runs it.

At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.

GPUCards neededFree for cacheContext it holds
RTX 4090 24GB 4 6.7 GiB about 36,337 tokens
RTX 5090 32GB 4 35.5 GiB about 193,624 tokens
L40S 48GB 2 11.7 GiB about 63,644 tokens
A100 80GB 2 69.3 GiB about 378,217 tokens
H100 80GB 2 69.3 GiB about 378,217 tokens
RTX PRO 6000 96GB 1 14.2 GiB about 77,297 tokens
H200 141GB 1 54.7 GiB about 298,481 tokens

Serving

Running it yourself.

A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.

vllm serve IFM/K2-Horizon-MoVA-36B-A4B \
  --max-model-len 32768 \
  --gpu-memory-utilization 0.90

Questions

The things people ask about this model.

How much GPU memory does K2-Horizon-MoVA-36B-A4B need?

The weights take about 70 GiB at the released precision. That fits on one RTX PRO 6000 96GB, which leaves roughly 14 GiB for the key-value cache.

What licence does K2-Horizon-MoVA-36B-A4B use?

The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.

How long a context does it support?

The configuration allows 524,288 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.

Can I use it through an API instead of hosting it?

This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.

Also in the library

Related models.

IFM/K2-Horizon-375B-A23B

K2-Horizon-375B-A23B is an open-weight sparse MiE language model from IFM with 379 B parameters, 23 B active per token and a 512 …

IFM/K2-Horizon-7B

K2-Horizon-7B is a 9-billion-parameter decoder-only language model with a 524k token context window, released under Apache-2.0.

IFM/K2-Horizon-3.7B

K2-Horizon-3.7B is a 3.7 billion-parameter dense decoder-only model with a 512 K token context, released openly by IFM.

IFM/K2-Horizon-32B

K2-Horizon-32B-Stage1 is a 32-billion-parameter dense decoder-only model released by IFM for benchmarking agentic, coding and rea…

Hosting

Want this model on a dedicated EU GPU?

Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.