MichelRosselli/ ternary-bonsai

27 Downloads Updated 4 days ago

Ternary Bonsai: 1.58-bit Language Models at 8B, 4B, and 1.7B Scale.

tools thinking

ollama run MichelRosselli/ternary-bonsai

curl http://localhost:11434/api/chat \
  -d '{
    "model": "MichelRosselli/ternary-bonsai",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='MichelRosselli/ternary-bonsai',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'MichelRosselli/ternary-bonsai',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code

Claude Code ollama launch claude --model MichelRosselli/ternary-bonsai

Codex App

Codex App ollama launch codex-app --model MichelRosselli/ternary-bonsai

OpenClaw

OpenClaw ollama launch openclaw --model MichelRosselli/ternary-bonsai

Hermes Agent

Hermes Agent ollama launch hermes --model MichelRosselli/ternary-bonsai

Codex

Codex ollama launch codex --model MichelRosselli/ternary-bonsai

OpenCode

OpenCode ollama launch opencode --model MichelRosselli/ternary-bonsai

Models

Name

4 models

Size / Usage

Context

Input

ternary-bonsai:latest

16GB · 64K context window · Text · 4 days ago

ternary-bonsai:latest

16GB

64K

Text

ternary-bonsai:1.7b-f16

3.4GB · 32K context window · Text · 4 days ago

ternary-bonsai:1.7b-f16

3.4GB

32K

Text

ternary-bonsai:4b-f16

8.0GB · 32K context window · Text · 4 days ago

ternary-bonsai:4b-f16

8.0GB

32K

Text

ternary-bonsai:8b-f16

16GB · 64K context window · Text · 4 days ago

ternary-bonsai:8b-f16

16GB

64K

Text

Readme

Ternary Bonsai

This is an unofficial Ollama upload of Prism ML’s Ternary Bonsai model family. These Ollama tags use the F16 GGUF variants for compatibility, with HF-aligned chat templates, long context, and tool-call support.

Tags

Tag	Size	Context	Notes
`latest`	8B F16	65,536	Alias of `8b-f16`
`8b-f16`	8B F16	65,536	Strongest/default variant
`4b-f16`	4B F16	32,768	Middle-size variant
`1.7b-f16`	1.7B F16	32,768	Smallest variant

Usage

ollama run MichelRosselli/ternary-bonsai

ollama run MichelRosselli/ternary-bonsai:1.7b-f16
ollama run MichelRosselli/ternary-bonsai:4b-f16
ollama run MichelRosselli/ternary-bonsai:8b-f16

Notes

The original Ternary Bonsai release explores ternary / 1.58-bit models.
These Ollama uploads use the F16 GGUF variants; Q2_0 ternary-packed GGUF is not supported here yet.
Tool calls are supported through PARSER qwen3.
Thinking text may be emitted as <think>...</think>, but Ollama does not expose it as a separate structured thinking field with the selected parser.
The chat template was validated against the Hugging Face template; remaining differences are limited to compact JSON formatting for tool schemas/calls.

Attribution

All credit for the original model belongs to Prism ML.

Original repositories:

License: Apache 2.0.

# Ternary Bonsai

This is an unofficial Ollama upload of Prism ML's Ternary Bonsai model family.
These Ollama tags use the F16 GGUF variants for compatibility, with
HF-aligned chat templates, long context, and tool-call support.

## Tags

| Tag | Size | Context | Notes |
| --- | ---: | ---: | --- |
| `latest` | 8B F16 | 65,536 | Alias of `8b-f16` |
| `8b-f16` | 8B F16 | 65,536 | Strongest/default variant |
| `4b-f16` | 4B F16 | 32,768 | Middle-size variant |
| `1.7b-f16` | 1.7B F16 | 32,768 | Smallest variant |

## Usage

```bash
ollama run MichelRosselli/ternary-bonsai
```

```bash
ollama run MichelRosselli/ternary-bonsai:1.7b-f16
ollama run MichelRosselli/ternary-bonsai:4b-f16
ollama run MichelRosselli/ternary-bonsai:8b-f16
```

## Notes

- The original Ternary Bonsai release explores ternary / 1.58-bit models.
- These Ollama uploads use the F16 GGUF variants; Q2_0 ternary-packed GGUF is
  not supported here yet.
- Tool calls are supported through `PARSER qwen3`.
- Thinking text may be emitted as `<think>...</think>`, but Ollama does not
  expose it as a separate structured thinking field with the selected parser.
- The chat template was validated against the Hugging Face template; remaining
  differences are limited to compact JSON formatting for tool schemas/calls.

## Attribution

All credit for the original model belongs to Prism ML.

Original repositories:

- https://huggingface.co/prism-ml/Ternary-Bonsai-1.7B-gguf
- https://huggingface.co/prism-ml/Ternary-Bonsai-4B-gguf
- https://huggingface.co/prism-ml/Ternary-Bonsai-8B-gguf
- https://huggingface.co/prism-ml/Ternary-Bonsai-1.7B-unpacked
- https://huggingface.co/prism-ml/Ternary-Bonsai-4B-unpacked
- https://huggingface.co/prism-ml/Ternary-Bonsai-8B-unpacked

License: Apache 2.0.

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)