---
title: granite-4.1-8b-FlintStones-V1-mxfp8-mlx
canonical_url: "https://www.modelscope.cn/models/nightmedia/granite-4.1-8b-FlintStones-V1-mxfp8-mlx"
md_url: "https://www.modelscope.cn/models/nightmedia/granite-4.1-8b-FlintStones-V1-mxfp8-mlx.md"
repository: nightmedia/granite-4.1-8b-FlintStones-V1-mxfp8-mlx
last_updated: 2026-05-03
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - granite
architectures:
  - GraniteForCausalLM
base_model:
  - ibm-granite/granite-4.1-8b
  - DavidAU/granite-4.1-8b-FlintStones-V1
base_model_relation: merge
parameters: 2.5B
tensor_type:
  - BF16
  - U8
  - U32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 12
stars: 0
tags:
  - finetune
  - unsloth
  - granite-4.1
  - reasoning
  - thinking
  - mxfp8
  - mergekit
  - merge
  - mlx
---

# granite-4.1-8b-FlintStones-V1-mxfp8-mlx

> granite-4.1-8b-FlintStones-V1-mxfp8-mlx - nightmedia 在 ModelScope 开源的模型。granite-4.1-8b-FlintStones-V1-mxfp8-mlx

nightmedia/granite-4.1-8b-FlintStones-V1-mxfp8-mlx 是 ModelScope 魔搭社区上的 2.5B 参数text-generation模型，采用 apache-2.0 许可，基于 ibm-granite/granite-4.1-8b、DavidAU/granite-4.1-8b-FlintStones-V1 构建。

- **Repository**: nightmedia/granite-4.1-8b-FlintStones-V1-mxfp8-mlx
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 2.5B
- **Base model**: ibm-granite/granite-4.1-8b, DavidAU/granite-4.1-8b-FlintStones-V1
- **Tags**: finetune, unsloth, granite-4.1, reasoning, thinking, mxfp8, mergekit, merge, mlx
- **Downloads**: 12
- **Stars**: 0
- **Last updated**: 2026-05-03

Source: https://www.modelscope.cn/models/nightmedia/granite-4.1-8b-FlintStones-V1-mxfp8-mlx

---

# granite-4.1-8b-FlintStones-V1-mxfp8-mlx


![image](https://cdn-uploads.huggingface.co/production/uploads/67b0caceb06805a4370c44c5/e9KpYOE-6jn6miBJZL-aQ.png)

For those too young to get the reference

https://www.britannica.com/topic/The-Flintstones

Brainwaves
```brainwaves
         arc   arc/e boolq hswag obkqa piqa  wino
mxfp8    0.512,0.701,0.860

granite-4.1-8b
mxfp8    0.486,0.666,0.875,0.636,0.450,0.766,0.631
```

It's very happy-go-lucky on bulletpoints, that's a trademark of Polaris Alpha training. Once we have a Claude to merge it with, this model will go places.

For now, lots of bulletpoints. IBM is an expert in that anyway, so probably nobody would notice

-G


This model [granite-4.1-8b-FlintStones-V1-mxfp8-mlx](https://huggingface.co/granite-4.1-8b-FlintStones-V1-mxfp8-mlx) was
converted to MLX format from [DavidAU/granite-4.1-8b-FlintStones-V1](https://huggingface.co/DavidAU/granite-4.1-8b-FlintStones-V1)
using mlx-lm version **0.31.3**.

## Use with mlx

```bash
pip install mlx-lm
```

```python
from mlx_lm import load, generate

model, tokenizer = load("granite-4.1-8b-FlintStones-V1-mxfp8-mlx")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True, return_dict=False,
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)
```