richardyoung/ hemmingway-1:Q6_K

160 Downloads Updated 2 weeks ago

Hemmingway-1 by Altworld: a 27B Qwen3.8 fine-tune that writes like a person. Everyday messages, emails and hard asks, without the preamble. Thinking and tools supported.

ollama run richardyoung/hemmingway-1:Q6_K

curl http://localhost:11434/api/chat \
  -d '{
    "model": "richardyoung/hemmingway-1:Q6_K",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='richardyoung/hemmingway-1:Q6_K',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'richardyoung/hemmingway-1:Q6_K',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 2 weeks ago

2 weeks ago

49de87221224 · 24GB

model

archqwen35

·

parameters27.3B

·

quantizationQ6_K

24GB

license

Hemmingway-1 by Altworld - Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.

382B

params

{ "num_ctx": 16384, "temperature": 1, "top_k": 20, "top_p": 0.95 }

58B

Readme

hemmingway-1

Hemmingway-1 by Altworld: a 27B Qwen3.8 fine-tune that writes like a person. Everyday messages, emails and hard asks, without the preamble. Thinking and tools supported.

Repackaged for Ollama from bartowski’s imatrix GGUFs. All credit to Altworld. Licensed CC BY-NC 4.0: free for non-commercial use with credit; commercial use needs an agreement with Altworld (luka@hemmingway.io).

Run

ollama run richardyoung/hemmingway-1

Tag	Size	Fits
`IQ3_M`	14.9 GB	16 GB GPU / 24 GB Mac
`IQ4_XS`	15.5 GB	24 GB GPU / 32 GB Mac
`latest`, `Q4_K_M`	17.4 GB	24 GB GPU / 32 GB Mac (recommended)
`Q5_K_M`	20.9 GB	24 GB GPU / 36 GB+ Mac
`Q6_K`	23.9 GB	32 GB GPU / 48 GB Mac
`Q8_0`	29.1 GB	48 GB GPU / 64 GB Mac (near-lossless)

Default context is 16K (native 262K); raise it with /set parameter num_ctx 32768.

Tips

Everyday writing (texts, emails, replies): turn thinking off (/set nothink, or "think": false in the API) for a fast, direct answer.
Stories and longer pieces: leave thinking on. It plans at length first and the prose is noticeably better.
Sampling follows Altworld’s defaults: temperature 1.0, top_k 20, top_p 0.95.

Benchmarks

Charts by Altworld, measured on the original BF16 weights. Quantized tags may score slightly lower. CommunicationBench and Human-Likeness are Altworld’s own blind benchmarks; EQ-Bench 4 is independent.

About

Built on Qwen3.8-27B. Per Altworld’s own blind head-to-head benchmarks it ranks first on CommunicationBench and Human-Likeness, and it places third on the public EQ-Bench 4. It is weaker at hostile storytelling and long story turns.

English-first. It can be wrong and still sound certain, so don’t use it for medical, legal or financial decisions.

Source: Altworld/Hemmingway-1 · GGUFs: bartowski/Altworld_Hemmingway-1-GGUF

Packaged for Ollama by Richard Young · DeepNeuro

# hemmingway-1

Hemmingway-1 by Altworld: a 27B Qwen3.8 fine-tune that writes like a person. Everyday messages, emails and hard asks, without the preamble. Thinking and tools supported.

> Repackaged for Ollama from bartowski's imatrix GGUFs. All credit to [Altworld](https://huggingface.co/Altworld/Hemmingway-1). Licensed **CC BY-NC 4.0**: free for non-commercial use with credit; commercial use needs an agreement with Altworld (luka@hemmingway.io).

## Run

```bash
ollama run richardyoung/hemmingway-1
```

| Tag                | Size    | Fits                                  |
| :----------------- | :------ | :------------------------------------ |
| `IQ3_M`            | 14.9 GB | 16 GB GPU / 24 GB Mac                 |
| `IQ4_XS`           | 15.5 GB | 24 GB GPU / 32 GB Mac                 |
| `latest`, `Q4_K_M` | 17.4 GB | 24 GB GPU / 32 GB Mac (recommended)   |
| `Q5_K_M`           | 20.9 GB | 24 GB GPU / 36 GB+ Mac                |
| `Q6_K`             | 23.9 GB | 32 GB GPU / 48 GB Mac                 |
| `Q8_0`             | 29.1 GB | 48 GB GPU / 64 GB Mac (near-lossless) |

Default context is 16K (native 262K); raise it with `/set parameter num_ctx 32768`.

## Tips

- **Everyday writing** (texts, emails, replies): turn thinking off (`/set nothink`, or `"think": false` in the API) for a fast, direct answer.
- **Stories and longer pieces**: leave thinking on. It plans at length first and the prose is noticeably better.
- Sampling follows Altworld's defaults: temperature 1.0, top_k 20, top_p 0.95.

## Benchmarks

*Charts by Altworld, measured on the original BF16 weights. Quantized tags may score slightly lower. CommunicationBench and Human-Likeness are Altworld's own blind benchmarks; EQ-Bench 4 is independent.*

![CommunicationBench](https://huggingface.co/Altworld/Hemmingway-1/resolve/main/charts/communicationbench.png)

![Human-Likeness](https://huggingface.co/Altworld/Hemmingway-1/resolve/main/charts/human-likeness.png)

![EQ-Bench 4](https://huggingface.co/Altworld/Hemmingway-1/resolve/main/charts/eqbench4.png)

## About

Built on Qwen3.8-27B. Per Altworld's own blind head-to-head benchmarks it ranks first on CommunicationBench and Human-Likeness, and it places third on the public EQ-Bench 4. It is weaker at hostile storytelling and long story turns.

English-first. It can be wrong and still sound certain, so don't use it for medical, legal or financial decisions.

**Source:** [Altworld/Hemmingway-1](https://huggingface.co/Altworld/Hemmingway-1) · **GGUFs:** [bartowski/Altworld_Hemmingway-1-GGUF](https://huggingface.co/bartowski/Altworld_Hemmingway-1-GGUF)

*Packaged for Ollama by [Richard Young](https://deepneuro.ai/richard) · DeepNeuro*

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