Skip to content
ModelCap

K2 Horizon MoVA 36B A4B

Official publisher releaseOpen weights

IFM·#93 of 350 on the current board

K2 Horizon MoVA 36B A4B is a language model from IFM. It ranks #93 of 350 on ModelCap's current board, with an Index score of 55.8, modeled from public metadata, not measured. Its score range of 30.5–81.1 spans positions #8–#232. Its weights are openly downloadable.

Figures as of 24 Oct 2026, 19:29 UTC · download the public dataset · methodology

ModelCap Index · Current board
55.8#93
Modeled from public metadata, not measured · range 30.5–81.1
Market Gravity
15.2Increased by 0.1
Secondary market signal

OpenRouter endpoint status could not be refreshed. This model is not treated as currently served until a successful observation arrives.

Input
Self-hosted
weights only · no listed API price
Output
Self-hosted
weights only · no listed API price
Context
—
tokens
Providers
0
endpoints

Providers

Uptime measured over the last 30 minutes

Current provider status is unavailable.

ModelCap Index · Current board

Methodology
55.8
Modeled from public metadata, not measured
Current board
#93 of 350
Score range
30.5–81.1
Rank range
#8–#232
Technical details
Public rank basis
Modeled · prior (architecture)
Evidence label
Modeled · prior · global-corpus-prior over 211 held-out anchors (72%); launch card against 2 resolved peers on 8 rows (28%); exceeds every named peer on 4 of 8 rows; optimism haircut 2.24 from cross-lab-probe; shrunk 9.1 toward the measured corpus
Index support
37.1% · limited
Identity binding
Exact catalogue product · ifm/IFM/K2-Horizon-MoVA-36B-A4B · aggregates disclosed benchmark configurations · endpoint configuration not claimed · artifact metadata IFM/K2-Horizon-MoVA-36B-A4B@e9ce717642b2074082875be233c02c729ba47960 (not the evaluation revision)
Index observations used
0
Qualified Index evidence point
55.8
How this position was produced · fused launch placement

No general-family board lists this model yet. Its point estimate is the precision-weighted mean of every channel that spoke for it; each channel’s share is the weight it carried. Independent measurements replace this placement as boards list the model.

Fused estimate
55.8 ± 19.8
Independent share
0%
Shrink toward corpus
-9.1
Peer ceiling share
50%
Optimism haircut
−2.2 points · cross lab probe, 5 probes
Method
precision-weighted-v1
Channels fused into this model’s launch placement
ChannelEstimateShare
Publisher corpus prior · not independent
global-corpus-prior over 211 held-out anchors
52.3 ± 23.372%
Launch card against resolved peers · not independent
launch card against 2 resolved peers on 8 rows
64.9 ± 37.428%
  • · Launch card row aa-lcr: inside the resolved peer range
  • · Launch card row aa-omniscience-non-hallucination: claims above every resolved peer
  • · Launch card row aa-omniscience: claims below every resolved peer
  • · Launch card row gpqa-diamond: inside the resolved peer range
  • · Launch card row hle-without-tools: claims above every resolved peer
  • · Launch card row scicode: inside the resolved peer range
  • · Launch card row tau-3-bench-banking: claims above every resolved peer
  • · Launch card row terminal-bench-2-1: claims above every resolved peer
Bayesian Evidence Score · measured evidence input only

BES normalizes admitted public benchmark evidence for measured Index rows. Its legacy score and observables prior do not define the public language rank.

Observed capability
—
Evidence status
No admitted measured evidence
General preference
—
Coding
—
Agents & tools
—
Reasoning
—
Evidence breadth
0%
Evidence coverage
0%
Market and catalogue signals · secondary, never the language Index rank
Usage (OpenRouter popularity)
—
Liquidity (providers × uptime)
—
Open reach (HF downloads)
74.1
Surface (context / tools / modalities)
0.0
Freshness
81.8
  • · HF 30d downloads 10983
  • · Catalogue surface
  • · First seen 2026-09-01

These adoption and deployment observations remain context only. They do not change this model's ModelCap Index score or rank.

The bounded 0–100 capability score blends each source’s competitive placement with its published achievement, then combines capability families. Evidence breadth adds a modest uncertainty adjustment; price and popularity are not part of benchmark-led rank.

Experimental capability estimate

Estimate policy
22.7
7.8–37.6 uncertainty interval
Excluded from rank
Confidence
76%
Training support
221 models · 129 lineages
Out-of-domain check
in domain
Feature coverage
71%
Lineage-held-out error
8.3 MAE

Experimental estimate. This layer predicts from admitted evidence and safe metadata only. It never enters ModelCap Score or rank.

Deployment readiness

Metadata index
47.9
partial · 52% metadata coverage
Not capability
Artifact reproducibility
85.0
Access & legal clarity
100.0
Deployability
8.0
Evaluation provenance
0.0

Not yet published: the base model it derives from, a current serving provider, measured provider uptime, published provider pricing, declared context and output limits, admitted benchmark results, results across several capability areas, results from more than one source, a confident match to its benchmark entries, pinned evaluation dataset versions.

A metadata completeness and deployability index—not a safety certification, quality grade, or production-readiness claim.

Market activity

Venue coverage
Market Gravity
15.2
OpenRouter weekly popularity
Not listed

Market Gravity uses OpenRouter’s full-catalogue popularity order. Sparse Vercel top-ten observations appear only when published and do not affect the score.

Arena preference

No matched Arena rating

This model is visible because it is available in the live catalogue, but it has no trustworthy community result yet. ModelCap does not infer a quality score from price, features, or market activity.

Market Gravity breakdown

Usage55%
0.0

OpenRouter weekly popularity across the full model catalogue.

Liquidity25%
0.0

Independent providers versus the model's open or closed cohort, weighted by uptime.

Open reach15%
74.1

Hugging Face 30-day downloads for open models; neutral for closed models.

Freshness5%
81.8

Time since first public availability, on a six-month half-life.

Specification

Model ID
IFM/K2-Horizon-MoVA-36B-A4B
Declared output cap
Unknown

OpenRouter catalogue or provider-endpoint declaration. Limits can vary by provider and prompt length; a model-wide maximum has not been independently verified. OpenRouter metadata

Inputs
Text
Outputs
Text
Cached input tokens
Not offered
First seen on OpenRouter
1 Sept 2026

Capabilities

  • Not supported: Reasoning
  • Not supported: Tool use
  • Not supported: Structured output
  • Not supported: Response format
  • Not supported: Moderated

Weights & access

Open weights. Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI.

Access
Open weights
Licence
apache-2.0
Pinned revision
e9ce717642b2
Parameters
37.4B
Architecture
K2HorizonForCausalLM
Model type
k2_horizon
Hugging Face downloads (30d)
11K
Downloads (all time)
30K
Likes
423
Repository updated
14 Oct 2026, 10:29 UTC

Alternatives to K2 Horizon MoVA 36B A4B, by the numbers

Ranked neighbours

Models within 3 places of #93 on the ModelCap Index.

Other ranked IFM models

The same publisher's other current models with a public Index position.

K2 Horizon MoVA 36B A4B: common questions

How does K2 Horizon MoVA 36B A4B rank among AI models?

K2 Horizon MoVA 36B A4B holds ModelCap Index position #93 of 350 ranked language models as of 24 October 2026, with an Index score of 55.8 (interval 30.5–81.1); evidence: modeled · prior. The Index combines public benchmark boards with published uncertainty; the methodology page explains the weighting.

How much does K2 Horizon MoVA 36B A4B cost per 1M tokens?

No live API price is listed for K2 Horizon MoVA 36B A4B in the current snapshot as of 24 October 2026 because the latest endpoint refresh was unavailable. ModelCap shows a price only when a provider currently lists one.

Is K2 Horizon MoVA 36B A4B open-weight?

Open weights (apache-2.0). Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI. The repository is IFM/K2-Horizon-MoVA-36B-A4B on Hugging Face.

Which benchmarks has K2 Horizon MoVA 36B A4B been evaluated on?

K2 Horizon MoVA 36B A4B's Index position rests on the evidence listed on this page; it has no result on the public boards ModelCap tracks as separate leaderboards as of 24 October 2026. The evidence panel names each source and its publication date.

When was K2 Horizon MoVA 36B A4B released?

K2 Horizon MoVA 36B A4B first appeared in the catalogue on 1 Sept 2026. It is the current version in its family. Rank and price movements since then are recorded on the site's changes feed.