moonshotai/Kimi-K2.5 · Discussions

#101 opened 18 days ago by

SaylorTwift

Support for video input and "4x temporal compression" with vLLM / SGLang

👍 2

#99 opened 18 days ago by

yzyy73

fp8 fmt for hopper ep deploy?

#98 opened about 1 month ago by

whybeyoung

Add SWE-Bench Verified evaluation results

#97 opened about 1 month ago by

Add MathArena evaluation result for hmmt/hmmt_feb_2026

#95 opened about 1 month ago by

JasperDekoninck

Very good model for simulating Baba Vanga

4

#94 opened about 1 month ago by

krustik

About Kimi Code

🚀 1

#93 opened about 1 month ago by

llMss12

clone this

#92 opened about 1 month ago by

givikjathan

fix AttributeError: 'MoonViT3dEncoder' object has no attribute 'use_deterministic_attn'

#91 opened about 1 month ago by

katuni4ka

Add Terminal-Bench evaluation result (43.2%)

#90 opened about 1 month ago by

burtenshaw

Possibility of a 10b model?

#89 opened about 1 month ago by

Crownelius

Add MathArena evaluation result for aime/aime_2026

#88 opened about 1 month ago by

JasperDekoninck

Does Kimi2.5 only score 72.57 on tau2-bench?

#87 opened about 2 months ago by

Alicia-Ross

All good

#86 opened about 2 months ago by

suprapto089

pathed_the_params.bash

#80 opened about 2 months ago by

Kjppmp

Can someone abliterate this model please

#79 opened about 2 months ago by deleted

BF16/FP8 version availability?

#77 opened about 2 months ago by

JFLOR274

Ai _chAt bot

#74 opened about 2 months ago by

vijaybarman3

The wave

#73 opened about 2 months ago by

Nk98James

Kimi K2.5 slow?

#72 opened about 2 months ago by

sebastienbo

Knowledge Cut-off

#71 opened about 2 months ago by

virtualhat

H200 - zero gpu

#70 opened about 2 months ago by

rahul7star

未正确识别工具调用

3

#69 opened about 2 months ago by

yeguangsuixing

age

➕ 1

#68 opened about 2 months ago by

DannySc

Model parameters confusion

#67 opened about 2 months ago by

Ujjwal-Tyagi

Add converted tokenizer (no trust_remote_code needed)

#66 opened about 2 months ago by

ArthurZ

first commit

#65 opened about 2 months ago by

pika19

success

#64 opened about 2 months ago by

pika19

多图输入的占位符或顺序？

4

#63 opened about 2 months ago by

noobimp

Value error, Model architectures ['KimiK25ForConditionalGeneration'] are not supported for now. Supported architectures:

#62 opened about 2 months ago by

jianyoulin

performance of quantized models

3

#61 opened about 2 months ago by

darvec

400 Bad Request when deploying Kimi-K2.5 with vllm-0.15.0

#60 opened about 2 months ago by

tuo02

hi, who are you? -> I am Claude.

6

#59 opened about 2 months ago by

Jeol

Model

#57 opened about 2 months ago by

Vickweb3

Navigation of hugging face

#55 opened about 2 months ago by

Xaviercrypt01

Model not generating </think> token .

➕ 6

3

#52 opened about 2 months ago by

pratiknarola

Add MMLU-Pro evaluation results

#51 opened about 2 months ago by

Add GPQA diamond evaluation results

#50 opened about 2 months ago by

Add HLE evaluation results (with tools)

🚀 1

#49 opened about 2 months ago by

The content of response from Kimi-K2.5 is empty when deploying with vllm-0.15.0.

5

#40 opened 2 months ago by

tuo02

AttributeError: 'CompressedLinear' object has no attribute 'weight'

#39 opened 2 months ago by

xbeark

Kimi K2.5 claims to be Claude, behaves like it, and even references Anthropic's Constitutional AI.

🤗👀 13

6

#38 opened 2 months ago by

ombd

Update chat template to use reasoning if reasoning_content is unavailable, which is deprecated