bartowski
Kwaipilot KAT Coder V2.5 Dev
Kwaipilot KAT Coder V2.5 Dev is a 34.7B mixture-of-experts qwen35moe model from bartowski, with about 6.3B parameters active per token. Configured context length is 262,144 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters34.7B / 6.3B active
Context256K
Attentionhybrid
Licenseapache-2.0
40 layers · 2 KV heads · head dim 256 · 30 linear-attention layers · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 64.6 GB | Hugging Face file bytes |
| Q8_0 | 34.4 GB | Hugging Face file bytes |
| Q6_K | 56.2 GB | Hugging Face file bytes |
| Q5_K_M | 23.3 GB | Hugging Face file bytes |
| Q5_K_S | 22.5 GB | Hugging Face file bytes |
| Q4_K_M | 19.9 GB | Hugging Face file bytes |
| Q4_K_S | 19.2 GB | Hugging Face file bytes |
| IQ4_NL | 18.5 GB | Hugging Face file bytes |
| IQ4_XS | 17.5 GB | Hugging Face file bytes |
| Q4_1 | 20.5 GB | Hugging Face file bytes |
| Q4_0 | 18.6 GB | Hugging Face file bytes |
| Q3_K_L | 15.7 GB | Hugging Face file bytes |
| Q3_K_M | 15.1 GB | Hugging Face file bytes |
| IQ3_M | 15.7 GB | Hugging Face file bytes |
| Q3_K_S | 14.4 GB | Hugging Face file bytes |
| IQ3_XXS | 13.8 GB | Hugging Face file bytes |
| Q2_K | 24.0 GB | Hugging Face file bytes |
| IQ2_S | 10.3 GB | Hugging Face file bytes |
| IQ2_XXS | 9.11 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Expert offload | Q3_K_L | 49–66 |
| GeForce RTX 3060 12GB | Tight fit | IQ2_XXS | 102–139 |
| GeForce RTX 4060 Ti 16GB | Tight fit | IQ3_XXS | 57–78 |
| GeForce RTX 4090 | Tight fit | Q4_K_M | 137–187 |
| GeForce RTX 5090 | Runs | Q5_K_M | 210–286 |
| Apple M4 Pro | Tight fit | Q3_K_L | 47–63 |
| Apple M5 | Tight fit | IQ2_S | 40–55 |
| CPU only · DDR5-5600 | Runs on CPU | Q5_K_M | 10–14 |
Smallest devices that fit it fully
- GeForce RTX 2080 Ti — Tight fit, IQ2_XXS, 175–238 tok/s
- GeForce GTX 1080 Ti — Tight fit, IQ2_XXS, 137–187 tok/s
- GeForce RTX 5070 — Tight fit, IQ2_XXS, 191–260 tok/s
- GeForce RTX 4070 Ti — Tight fit, IQ2_XXS, 143–195 tok/s
- GeForce RTX 4070 Super — Tight fit, IQ2_XXS, 143–195 tok/s
- GeForce RTX 4070 — Tight fit, IQ2_XXS, 143–195 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (IQ3_XXS).
llama.cpp
llama-server -hf bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF:IQ3_XXS -c 4096 -ngl 40
No matching name in the Ollama library snapshot.
LM Studio
lms get bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF # Then load quant IQ3_XXS