Bonsai 2 27B

Ternary · llama.cpp · 128K context · RTX PRO 4500 Blackwell · 32 GB

Prism ML's ternary Bonsai 2, built on Qwen3.6-27B: a 27B model packed small enough for 16 GB, with vision, tools and thinking.

– tok/sdecode

– tok/sprefill

128Kcontext window

Earlier checkSep 22, 2026

Run it

Follow the steps below to download and run this recipe. Omarchy Local AI lists the recipes it can launch on your hardware. The server then answers on http://localhost:8000/v1.

1 Download the weights

hf download prism-ml/Ternary-Bonsai-2-27B-gguf \
  --revision 6ed5e12bf84b7a63069882c91dd9e9218647d17b

2 Start the server

docker run --rm \
  --gpus all \
  -p 8000:8080 \
  --ipc host \
  --shm-size 16g \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  --entrypoint /opt/llama/llama-server \
  ghcr.io/0xsero/deploy-llamacpp-bonsai-ptq1-rtx-2000-ada-16gb@sha256:9a646183bf0dd199504bea9b6f47d9ec2aa07a3bdf1a4827382f449e94292f02 \
  /opt/llama/llama-server \
  --model /opt/models/Ternary-Bonsai-2-27B/Ternary-Bonsai-2-27B-PTQ1_0.gguf \
  --mmproj /opt/models/Ternary-Bonsai-2-27B/Ternary-Bonsai-2-27B-mmproj-Q8_0.gguf \
  --alias Ternary-Bonsai-2-27B-PTQ1_0 \
  --host 0.0.0.0 \
  --port 8080 \
  --ctx-size 131072 \
  --cache-type-k q4_0 \
  --cache-type-v q4_0 \
  --n-gpu-layers 99 \
  --jinja

What it passed

✓ loadLoads and serves within an hour

✓ chatAnswers a plain question and stops on its own

✓ reasoningThinks separately and gets 17 × 23 right

✓ toolsCalls a tool with the right arguments and uses the result

– contextRecalls a code buried in a prompt that fills 85% of the window

– speedDecodes at 15 tok/s or more

Passed the older check (loads and chats); a full six-check run is pending.

Details

Weights: prism-ml/Ternary-Bonsai-2-27B-gguf @ 6ed5e12bf8 ›
Image: ghcr.io/0xsero/deploy-llamacpp-bonsai-ptq1-rtx-2000-ada-16gb@sha256:9a646183bf0dd199504bea9b6f47d9ec2aa07a3bdf1a4827382f449e94292f02
Launch: llama.cpp-bonsai-2-27b-ptq1-0-128k-v3 ›
Recipe file: registry/recipes/nvidia/rtx-pro-4500-blackwell-32gb/bonsai-2-27b.llama.cpp.128k.json ›
Model released: Apr 21, 2026
Tested: earlier acceptance