---
title: Mellum2-12B-A2.5B-Thinking-mxfp4
canonical_url: "https://www.modelscope.cn/models/mlx-community/Mellum2-12B-A2.5B-Thinking-mxfp4"
md_url: "https://www.modelscope.cn/models/mlx-community/Mellum2-12B-A2.5B-Thinking-mxfp4.md"
repository: mlx-community/Mellum2-12B-A2.5B-Thinking-mxfp4
last_updated: 2026-08-12
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - mellum
architectures:
  - MellumForCausalLM
base_model:
  - JetBrains/Mellum2-12B-A2.5B-Thinking
base_model_relation: quantized
parameters: 2.3B
tensor_type:
  - U32
  - U8
  - BF16
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
downloads: 16
stars: 0
tags:
  - mlx
  - mellum
  - moe
  - thinking
  - mxfp4
---

# Mellum2-12B-A2.5B-Thinking-mxfp4

> Mellum2-12B-A2.5B-Thinking-mxfp4 - mlx-community 在 ModelScope 开源的模型。Mellum2 12B A2.5B Thinking - MXFP4 MLX

mlx-community/Mellum2-12B-A2.5B-Thinking-mxfp4 是 ModelScope 魔搭社区上的 2.3B 参数text-generation模型，采用 apache-2.0 许可，基于 JetBrains/Mellum2-12B-A2.5B-Thinking 构建。

- **Repository**: mlx-community/Mellum2-12B-A2.5B-Thinking-mxfp4
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 2.3B
- **Base model**: JetBrains/Mellum2-12B-A2.5B-Thinking
- **Tags**: mlx, mellum, moe, thinking, mxfp4
- **Downloads**: 16
- **Stars**: 0
- **Last updated**: 2026-08-12

Source: https://www.modelscope.cn/models/mlx-community/Mellum2-12B-A2.5B-Thinking-mxfp4

---

# Mellum2 12B A2.5B Thinking - MXFP4 MLX

This is an MXFP4 MLX quantization of
[JetBrains/Mellum2-12B-A2.5B-Thinking](https://huggingface.co/JetBrains/Mellum2-12B-A2.5B-Thinking).

Mellum2 Thinking is a reasoning-augmented Mixture-of-Experts assistant model.
It has 64 experts with 8 active per token, a 131,072-token context window, and
emits reasoning in `<think>...</think>` blocks before the final answer.

## Conversion details

- Source: `JetBrains/Mellum2-12B-A2.5B-Thinking`
- Format: MLX safetensors
- Quantization: MXFP4, 4 bits, group size 32
- License: Apache-2.0
- EOS token: `<|im_end|>` (token ID 28)

The upstream `config.json` and `generation_config.json` identify token ID 0 as
the EOS token, while the tokenizer identifies `<|im_end|>` (ID 28) as EOS.
This conversion uses token ID 28 so MLX generation stops at the end of the
assistant turn.

## Usage

```bash
pip install -U mlx-lm

mlx_lm.chat \
  --model mlx-community/Mellum2-12B-A2.5B-Thinking-mxfp4 \
  --max-tokens 8192 \
  --temp 0.6 \
  --top-p 0.95
```

## Benchmarks

Benchmarked with oMLX 0.5.7 and MLX-LM 0.31.3 on an Apple M5 Max MacBook
Pro with 128 GB unified memory, using the built-in Python-code workload and
128 generated tokens. Results vary with hardware, runtime versions, context
length, and generation settings.

| Prompt tokens | TTFT | Prefill | Generation | End-to-end | Peak memory |
| ---: | ---: | ---: | ---: | ---: | ---: |
| 4,096 | 963.4 ms | 4,251.6 tok/s | 172.3 tok/s | 1.7 s | 6.80 GB |
| 16,384 | 5,739.5 ms | 2,854.6 tok/s | 100.9 tok/s | 7.0 s | 7.04 GB |
| 32,768 | 17,226.8 ms | 1,902.2 tok/s | 101.1 tok/s | 18.5 s | 7.42 GB |
| 65,536 | 39,122.4 ms | 1,675.2 tok/s | 73.3 tok/s | 40.9 s | 8.12 GB |
| 131,072 | 83,223.6 ms | 1,574.9 tok/s | 42.2 tok/s | 86.3 s | 9.54 GB |

At a 4,096-token prompt, batch generation achieved 226.6, 183.6, and 220.6
aggregate generation tok/s at 2, 4, and 8 concurrent requests, respectively,
compared with 172.3 tok/s for one request.

## Model provenance

For the original model card, training details, benchmark results, and usage
guidance, see the
[upstream JetBrains checkpoint](https://huggingface.co/JetBrains/Mellum2-12B-A2.5B-Thinking).