---
title: lift-bf16
canonical_url: "https://www.modelscope.cn/models/mlx-community/lift-bf16"
md_url: "https://www.modelscope.cn/models/mlx-community/lift-bf16.md"
repository: mlx-community/lift-bf16
last_updated: 2026-06-28
license: openrail
pipeline_tag: image-text-to-text
tasks:
  - image-text-to-text
model_type:
  - qwen3_5
architectures:
  - Qwen3_5ForConditionalGeneration
base_model:
  - datalab-to/lift
base_model_relation: finetune
parameters: 9.4B
tensor_type:
  - BF16
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 36
stars: 0
tags:
  - pdf
  - extraction
  - structured-data
  - json
  - mlx
  - mlx-vlm
  - vlm
  - bf16
---

# lift-bf16

> lift-bf16 - mlx-community 在 ModelScope 开源的模型。MLX conversion of datalab-to/lift — a 9B qwen35 vision-language model for structured extraction (PDF / image → schema-constrained JSON).

mlx-community/lift-bf16 是 ModelScope 魔搭社区上的 9.4B 参数image-text-to-text模型，采用 openrail 许可，基于 datalab-to/lift 构建。

- **Repository**: mlx-community/lift-bf16
- **License**: openrail
- **Tasks**: image-text-to-text
- **Parameters**: 9.4B
- **Base model**: datalab-to/lift
- **Tags**: pdf, extraction, structured-data, json, mlx, mlx-vlm, vlm, bf16
- **Downloads**: 36
- **Stars**: 0
- **Last updated**: 2026-06-28

Source: https://www.modelscope.cn/models/mlx-community/lift-bf16

---

# lift-bf16

MLX conversion of [**datalab-to/lift**](https://huggingface.co/datalab-to/lift) — a 9B `qwen3_5`
vision-language model for **structured extraction** (PDF / image → schema-constrained JSON).

This repo is the **bf16** full-precision conversion — no quantization, the source weights re-saved in MLX format. Runs on Apple Silicon via [mlx-vlm](https://github.com/Blaizzy/mlx-vlm). Not affiliated with
Datalab — a community conversion of their openly released weights.

## Variants

| Repo | Method | ≈bpw | Size | Peak RAM\* | Gen\* |
|---|---|---|---|---|---|
| **lift-bf16 (this repo)** | **full bf16** | **16** | **18 GB** | **19.9 GB** | **31 t/s** |
| [lift-oQ8](https://huggingface.co/mlx-community/lift-oQ8) | oQ | ≈8.6 | 9.7 GB | 12.3 GB | 58 t/s |
| [lift-oQ6](https://huggingface.co/mlx-community/lift-oQ6) | oQ | ≈6 | 7.7 GB | 9.4 GB | 73 t/s |
| [lift-oQ5](https://huggingface.co/mlx-community/lift-oQ5) | oQ | ≈5 | 6.7 GB | 8.4 GB | 83 t/s |
| [lift-oQ4](https://huggingface.co/mlx-community/lift-oQ4) | oQ | ≈4.6 | 5.6 GB | 7.2 GB | 100 t/s |
| [lift-oQ3.5](https://huggingface.co/mlx-community/lift-oQ3.5) | oQ | ≈4.0 | 4.9 GB | 6.5 GB | 109 t/s |
| [lift-oQ3](https://huggingface.co/mlx-community/lift-oQ3) | oQ | ≈3.5 | 4.6 GB | 6.2 GB | 119 t/s |

\* Peak RAM and generation speed measured on a single-image invoice extraction on a Macbook Pro M5 Max 128GB 40 GPU. Indicative, not a benchmark.

## Usage

### Generate (CLI)

```bash
uvx --from mlx-vlm mlx_vlm.generate \
  --model mlx-community/lift-bf16 \
  --image invoice.png \
  --prompt "Extract the invoice as JSON." \
  --max-tokens 800
```

### OpenAI-compatible server + structured outputs

lift is built for **schema-constrained** extraction. `mlx_vlm.server` enforces a JSON Schema at decode
time (via `llguidance`), so output is guaranteed valid and well-typed.

```bash
uvx --from mlx-vlm mlx_vlm.server --model mlx-community/lift-bf16 --port 8080
```

```python
import base64, json
from openai import OpenAI

client = OpenAI(base_url="http://127.0.0.1:8080/v1", api_key="local")
img = base64.b64encode(open("invoice.png", "rb").read()).decode()

schema = {
    "type": "object",
    "properties": {
        "invoice_number": {"type": "string"},
        "total": {"type": "number"},
        "line_items": {"type": "array", "items": {"type": "object", "properties": {
            "description": {"type": "string"}, "amount": {"type": "number"}}}},
    },
    "required": ["invoice_number", "total"],
}

resp = client.chat.completions.create(
    model="mlx-community/lift-bf16",   # the server lists your whole HF cache — name the model explicitly
    messages=[{"role": "user", "content": [
        {"type": "text", "text": "Extract this invoice."},
        {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{img}"}},
    ]}],
    response_format={"type": "json_schema", "json_schema": {"name": "invoice", "schema": schema}},
    temperature=0.0, max_tokens=800,
)
print(json.loads(resp.choices[0].message.content))
```

## Notes

- **eos fix applied.** `generation_config.json` here sets `eos_token_id: [248044, 248046]`. Upstream
  only sets `248044`, but the chat turn closes with `<|im_end|>` = `248046`; without this, MLX servers
  reading `generation_config` never stop and flood `<|im_end|>`. If you re-convert from the source, reapply this.
- **Quality.** Upstream FP `lift` (9B) scores 90.2% field / 20.9% full-document on Datalab's 225-doc
  benchmark. Every variant in this set extracted a simple test invoice correctly — that only rules out
  collapse, it does **not** rank them. Lower bit-widths may degrade on harder/adversarial documents; these
  were not re-benchmarked at scale.

## License

Code Apache-2.0; weights under a modified OpenRAIL-M (free for research, personal use, and startups under
$5M; not for use competitive with Datalab's API). See the [base model](https://huggingface.co/datalab-to/lift).