← Back to Model Radar
#19 on the radar · Vision

Kimi-K3

unsloth · image-text-to-text
FreshFresh (<14d)Vision
▲ 127Trending score
410Downloads
♥ 128Likes
todayLast updated
#19of 150 tracked models
top 13% by trending score
Model IDunsloth/Kimi-K3
Authorunsloth
Taskimage-text-to-text
CategoryVision
Licenseother
Parameters2780B
Trending score127
Downloads410
Likes128
Radar rank#19
HealthFresh (<14d)
Freshnesstoday (0.2 days)
Last modified2026-07-28T06:38:35.000Z

Kimi-K3 is a image-text-to-text model in the Vision category, published by unsloth. These figures are Hugging Face's own signals — downloads, likes and trending score — captured by Model Radar and refreshed daily. For weights, model card and usage, open it on Hugging Face.

Surge analysis

why it's on the radar

Hugging Face publishes a single trending score; it doesn't expose the formula. So Model Radar reads the surge from the four public signals that move with it — momentum, adoption, community and recency — each normalized against the tracked field, then weighted into a 0–100 composite. The weights are ours, and shown below. No hidden inputs.

54Radar surge index / 100
Momentum 55 / 100 · 45% weight
HF trending score 127, log-scaled vs the field's peak of 6913.
Adoption 31 / 100 · 25% weight
410 downloads, log-scaled vs the field's peak.
Community 51 / 100 · 18% weight
128 likes, log-scaled vs the field's peak.
Recency 99 / 100 · 12% weight
Last updated today — newer weights higher.

Radar position over time

movement

Where Kimi-K3 sits on the radar, tracked daily — new to the radar since the prior run. position movement fills in as the radar refreshes daily.

#19Current rank
#19Best rank
NEWSince prior run

In context

vs. category peers

How this model stacks up against the other 41 Vision models on the radar. The filled bar is this model; the tick is the category-peer median. Downloads and likes use a log scale (they span orders of magnitude).

Trending score▲ 127 · peer median ▲ 46
Downloads410 · peer median 341k
Likes♥ 128 · peer median ♥ 770
Freshnesstoday · peer median 8 days ago

Bars are normalized against the most extreme model in the whole radar, so a near-full bar means category-leading. Peer median = the middle model of the 41 in this category.

Recency

last update

Where this model's last update lands on a freshness scale — fresh (<14d), recent (14–60d), or older.

updated today
today14d60d1y+

Category peers

Vision · 41 models

Other Vision models ranked near this one — click through to compare.

📛 Embed this badge

rank badge

Show your live Model Radar rank in your README — it updates daily and links back here.

Model Radar rank badge for Kimi-K3

[![Model Radar rank](https://modelradar.kymatalabs.com/badge/unsloth-kimi-k3.svg)](https://modelradar.kymatalabs.com/m/unsloth-kimi-k3/)
<a href="https://modelradar.kymatalabs.com/m/unsloth-kimi-k3/"><img src="https://modelradar.kymatalabs.com/badge/unsloth-kimi-k3.svg" alt="Model Radar rank"></a>

Primary links

on hugging face

Go straight to the source — everything here links back to Hugging Face.