whatmodel.sh / KAT-Coder-V2.5-Dev (35B-A3B)

KAT-Coder-V2.5-Dev (35B-A3B)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

KAT-Coder-V2.5-Dev (35B-A3B) is an open-weights language model from Kuaishou (Kwaipilot), released July 2026. Its published size is 35B parameters (3B active per token) and a 262k-token context window. Licence: apache-2.0.

Catalogue identity notes

Kwaipilot (Kuaishou) open-weight release Kwaipilot/KAT-Coder-V2.5-Dev @7be56fe (created 2026-07-23, Apache-2.0): 'Following the release of KAT-Coder-V2.5 in July, we are pleased to release the open-weight version KAT-Coder-V2.5-Dev'; post-trained from Qwen3.6-35B-A3B (SFT + RL); text-only (vision weights not shipped). Distinct from the API models KAT-Coder-Pro V1 (catalogued kat-coder-pro-v1) and KAT-Coder-V2/V2.5.

Model information sources

Release date ↗ · Weights licence ↗

Card: 'an MOE model with a total parameter count of 35B and 3B activated parameters'; safetensors metadata 34,660,610,688 (language-model weights only).

70.2±1.6#82 of 719 LLMsfrom 5 benchmark results across 1 source

$ how it compares

The LLMs ranked either side of KAT-Coder-V2.5-Dev (35B-A3B), out of 828 with a score.

#modelscorewhereRAMreleased
86Gemini 2.5 Pro Preview (06-05) Google70±2comparison only2025-06
87o3-pro OpenAI70±2api $35.00/M2025-06
88KAT-Coder-V2.5-Dev (35B-A3B) Kuaishou (Kwaipilot)70±2local~22GB2026-07
89Qwen3.6 Plus Alibaba70±1api $1.13/M2026-04
90Inkling Small (266B-A12B) Thinking Machines70±1api $0.64/M +local~86GB2026-07

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #84 of 300 · 88% of its score · 4 tests

Beats Grok 4.20 (0309 reasoning), GPT-5.1, Grok 4.3 here, though they rank higher overall. Trails Hy3 preview (295B-A21B), MiniMax-M2.7, Step 3.5 Flash (196B-A11B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Terminal-Bench 2.1 lab32.6% ✎▄······█▄▄█▄ beats 4/7very hard · avg 7.1%▇▇▇▇▇▇▇▇▇▇▇▇ 29%at its level
SWE-bench Verified lab69.4% ✎▄▄·····▄▄▄▄█ beats 7/7hard · avg 17%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level
SWE-bench Pro (split unspecified) lab46.0% ✎▄···▄···▄▄██ beats 7/7very hard · avg 6.5%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level
SWE-bench Multilingual lab63.0% ✎▄·▄·····█▄█▄ beats 7/7hard · avg 18%▇▇▇▇▇ 13%at its level
coding▁▁▁▃▆██▇▆▄▃▁ #65 of 458 · 70% of its score · 4 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-5.2, GPT-5 here, though they rank higher overall. Trails MiMo-V2.5-Pro (1T-A42B), Qwen3.6 Plus here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Verified lab69.4% ✎▄▄·····▄▄▄▄█ beats 7/7hard · avg 17%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level
SWE-bench Pro (split unspecified) lab46.0% ✎▄···▄···▄▄██ beats 7/7very hard · avg 6.5%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level
SWE-bench Multilingual lab63.0% ✎▄·▄·····█▄█▄ beats 7/7hard · avg 18%▇▇▇▇▇ 13%at its level
SciCode (level unspecified) lab44.2% ✎▄···▄▄█▄·▄·▄ beats 7/7very hard · avg 13%▇▇▇▇ 11%at its level

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run KAT-Coder-V2.5-Dev (35B-A3B) · needs ~21.39 GB · from huggingface.co/bartowski
    chat in the terminal
    llama cli -hf bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run KAT-Coder-V2.5-Dev (35B-A3B) · needs ~21.39 GB · from huggingface.co/bartowski

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run KAT-Coder-V2.5-Dev (35B-A3B) · needs ~21.39 GB · from huggingface.co/bartowski
    chat in the terminal
    ollama run hf.co/bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF:Q4_K_M

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run KAT-Coder-V2.5-Dev (35B-A3B) · needs ~19.5 GB, Mac only · from huggingface.co/majentik
    chat in the terminal
    mlx_lm.chat --model majentik/KAT-Coder-V2.5-Dev-MLX-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model majentik/KAT-Coder-V2.5-Dev-MLX-4bit

Official weights and documentation: huggingface.co/Kwaipilot ↗

$ how the score was worked out · Kwaipilot/KAT-Coder-V2.5-Dev… 99%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Kwaipilot/KAT-Coder-V2.5-Dev model card · lab-reported5▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 99%Apache-2.0
prior · pull toward the middle▇ 1%
weighted average= 70.2

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json