freakyskittle / kimi-k2.7-code-GGUF

Kimi K2.7 Code GGUF

GGUF conversions of moonshotai/Kimi-K2.7-Code .

This repository contains full-quality BF16 GGUF shards and low-bit GGUF quantizations for llama.cpp-compatible runtimes.

Available files

How to load split GGUF files

For split GGUF variants, point llama.cpp at the first shard. llama.cpp will discover the rest automatically when the shards are in the same directory.

Example for BF16:

llama-cli -m BF16/kimi-k2.7-code-BF16-00001-of-00061.gguf -p "Write a Python function for quicksort."

When Q4_K_M split shards are uploaded, use the same pattern with shard 00001 from the Q4_K_M/ directory.

Quantization notes

• BF16 was converted from the original SafeTensors using llama.cpp convert_hf_to_gguf.py with BF16 output.
• TQ1_0 and TQ2_0 are llama.cpp ternary low-bit formats.
• IQ1_S was not produced because llama.cpp requires an importance matrix for that quantization.
• Q4_K_M is the standard high-quality GGUF 4-bit quantization. It must be split for upload because the single-file output is over Hugging Face’s 500GB individual file limit.

License

See LICENSE . This model uses Moonshot AI’s Modified MIT License for Kimi K2.7 Code.

Attribution

Base model by Moonshot AI: moonshotai/Kimi-K2.7-Code .