moonshotai / image-text-to-text updated 3 months ago

Kimi-K2.5

Name: Kimi-K2.5
Author: moonshotai

- 2026.1.29: - The default system prompt might cause confusion to users and unexpected behaviours, so we remove it. - The token is incorrect; it has been replaced with in the chat template.

Params

1,058.6 B

Context

—

Downloads 30d

966 K

Likes

2,860

Licence: other Not gated SAFETENSORS View on Hugging Face ↗

Download history

daily snapshots · 10 days

987 K966 K

Jul 28Jul 31Aug 3Aug 6

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

File	Quant	Size	Est. VRAM	Verdict on RTX 4090 · 24 GB
model.safetensors	i32	595.2 GB	814.0 GB	❌ Won’t fit

Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Specifications

Architecture: KimiK25ForConditionalGeneration
Parameters: 1,058.6 B
Tensor type: I32
Licence: other
First seen on the Hub: 2026-01-01
Training datasets: undisclosed
Added to our catalog: 2026-07-28

Family

Base model and the most-downloaded derivatives in the catalog.

Derivative Kimi-K2.5-NVFP4 nvidia 298 K Derivative Kimi-K2.5 mlx-community 188 K

Compare with

Qwen3.5-9B 12.4 M Qwen2.5-VL-7B-Instruct 9.3 M Qwen2.5-VL-3B-Instruct 8.2 M Qwen3.6-27B 7.0 M Qwen3.5-4B 6.7 M Qwen3-VL-8B-Instruct 4.9 M

Appears in

Best open coding models Commercial-use text models Models that fit in 12 GB