---

license: apache-2.0

base_model: decent-jawfish/bonsai-2-27b-mtp

base_model_relation: quantized

library_name: gguf

pipeline_tag: text-generation

tags:

- gguf

- qwen3.5

- mtp

---

Bonsai-2-27B PQ2_0 MTP GGUF

This repository publishes the local file Bonsai-2-27B-PQ2_0-MTP.gguf (7,557,178,656 bytes) from decent-jawfish/bonsai-2-27b-mtp. That model card reports Apache-2.0. groxaxo did not quantize this file.

The GGUF header has no license field. It records architecture qwen35, size label 27B, basename folded, and general.file_type 141. The filename identifies the PQ2_0 MTP quant.

A 27 September 2026 bench note records SHA-256 78df4279d40ebebdccfd2dae0e9d4847afee52e94f48f3542ae9437220dbd847 for this filename. The upload checks that hash before publishing. SHA256SUMS is the hash of the uploaded bytes.

This file was benchmarked on another machine on 27 September 2026. That run was not repeated for this upload. The llama.cpp command recorded for that bench was:

llama-server -m Bonsai-2-27B-PQ2_0-MTP.gguf --alias bonsai-27b-pq2-mtp --jinja \
  -ngl 999 -fa on -c 16384 -np 1 -ctk q8_0 -ctv q8_0 \
  --temp 1.0 --top-p 0.95 --top-k 20 --min-p 0.05 \
  --spec-type draft-mtp --spec-draft-n-max 2

The measured scores stay in that bench write-up. They are not a new result for this repository.