This model is a fine-tuned version of Qwen/Qwen3.8-27B on the on the Omni-Edu-70K dataset. The following hyperparameters were used during training: - learningrate: 5e-06 - trainbatchsize: 1 - evalbatchsize: 8 - distributedtype: multi-GPU - numdevices: 16 - gradientaccumulationsteps: 8 - totaltrainbatchsize: 128 - totalevalbatchsize: 128 - lrschedulertype: cosine - lrschedulerwarmupsteps: 0.1 - numepochs: 3.0 - Transformers 5.2.0 - Pytorch 2.10.0 - Datasets 4.0.0 - Tokenizers 0.22.2
This Hub repository contains a HuggingFace's transformers implementation of Florence-2 model from Microsoft. Florence-2 is an advanced vision foundation model that uses a prompt-based approach to handle a wide range of vision and vision-language tasks.
Parameters232M
Context1,024
Weights927.6 MB
Licensemit
AccessOpen weights
Monthly Downloads3M
Runs On
What it takes to serve Florence-2-base (232M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 0.5 GB | 0.6 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.2 GB | 0.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.1 GB | 0.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
SAVRN's Notes on Florence-2-base
Hand it an image and a short text prompt and it returns text; Microsoft's examples are captioning, object detection and segmentation. At 232M parameters the 16-bit weights are 0.5 GB and the run needs 0.6 GB; 8-bit and 4-bit trim that to 0.3 GB and 0.1 GB. Our Index prices the cheapest qualifying card, a single 192 GB MI300X, at $1.85 an hour on-demand, which this would leave nearly empty, so the question is what else shares it.
MIT is the license: commercial use, modification and redistribution, with the copyright and permission notices kept in the package and nothing else to pass downstream. Access is open. Check the context and the revision. Context is 1,024 tokens: room for a task prompt and a caption back, not long text. The files were last updated 2025-08-04, over a year after the 2024-06-15 release, so pin the revision you tested.
Model Card
By Microsoft, published under mit, revision 5ca5edf5bd01.
Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
Model Summary
This Hub repository contains a HuggingFace's transformers implementation of Florence-2 model from Microsoft.
Florence-2 is an advanced vision foundation model that uses a prompt-based approach to handle a wide range of vision and vision-language tasks. Florence-2 can interpret simple text prompts to perform tasks like captioning, object detection, and segmentation. It leverages our FLD-5B dataset, containing 5.4 billion annotations across 126 million images, to master multi-task learning. The model's sequence-to-sequence architecture enables it to excel in both zero-shot and fine-tuned settings, proving to be a competitive vision foundation model.
Resources and Technical Documentation: + Florence-2 technical report. + Jupyter Notebook for inference and visualization of Florence-2-large model
Configuration
- Architecture
- Florence2ForConditionalGeneration
- Context length (tokens)
- 1,024
- Layers
- 6
- Vocabulary size
- 51,289
- Stored precision
- float16
- Model type
- florence2
Identity and Version
- Repository
- microsoft/Florence-2-base
- Publisher
- Microsoft
- Task
- Image and text to text
- Modality
- Image and text
- Library
- transformers
- Parameters
- 232M parameters
- Languages
- Not stated by the source
- Revision
- 5ca5edf5bd017b9919c05d08aebef5e4c7ac3bac
- First published
- 2024-06-15
- Last updated
- 2025-08-04
Files and Weights
16 files, 930.3 MB in total. The weights are 2 files totalling 927.6 MB in bin, safetensors.
Weights2 files · 927.6 MB
Configuration5 files · 194.5 KB
Tokenizer3 files · 2.5 MB
Documentation5 files · 20.3 KB
Repository1 file · 1.6 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model.safetensors | Weights | 463.2 MB | 03075d2d2d2b |
| pytorch_model.bin | Weights | 464.4 MB | b480ac374593 |
| config.json | Configuration | 2.4 KB | — |
| configuration_florence2.py | Configuration | 15.1 KB | — |
| modeling_florence2.py | Configuration | 127.5 KB | — |
| preprocessor_config.json | Configuration | 806 B | — |
| processing_florence2.py | Configuration | 48.7 KB | — |
| CODE_OF_CONDUCT.md | Documentation | 444 B | — |
| LICENSE | Documentation | 1.1 KB | — |
| README.md | Documentation | 14.8 KB | — |
| SECURITY.md | Documentation | 2.7 KB | — |
| SUPPORT.md | Documentation | 1.2 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 1.4 MB | — |
| tokenizer_config.json | Tokenizer | 34 B | — |
| vocab.json | Tokenizer | 1.1 MB | — |
License and Download
- License
- mit
- Access
- Open weights, no gate
- Download size
- 927.6 MB
Download from Microsoft
Released by Microsoft through its official repository on Hugging Face. Read the license.
Built From
- Described by arXiv:2311.06242
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 927.6 MB |
| 16-bit | 0.5 GB |
| 8-bit | 0.2 GB |
| 4-bit | 0.1 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Compare Florence-2-base
Questions About Florence-2-base
How much GPU memory does Florence-2-base need?
About 0.6 GB at 16-bit and 0.1 GB at 4-bit: the weights (232M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Florence-2-base on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use Florence-2-base commercially?
Yes. Florence-2-base is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.
What is Florence-2-base's context length?
1,024 tokens, from the maximum position embeddings in its published configuration.
Similar Models
I built this quant because the ready-made FP4 file answered the wrong question. It was fast, but on my short WikiText-2 control it scored 6.4949 PPL. Plain Q40 scored 6.3798. The first higher-quality hybrid went too far the other way: good perplexity, 34.19 tok/s, and no comfortable room for 256K plus vision. This is the build that survived both gates. It is a 17.1 GB, 5.01 BPW mixed-precision GGUF of Qwen/Qwen3.8-27B. It keeps large, tolerant matrices in native NVFP4 and spends more bits on selected attention, Gated DeltaNet, and late FFN tensors. The trained MTP layer remains embedded in the same GGUF. This is not a fine-tune. I built the private calibration workload from 5,472 messages…
This is an uncensored version of Qwen/Qwen3.8-27B created with abliteration (see remove-refusals-with-transformers to know more about it). This is a crude, proof-of-concept implementation to remove refusals from an LLM model without using TransformerLens. The newly added Huihui-Qwen3.8-27B-abliterated-GSQ-RCO series come from ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF. Only layers 23 to 51 have been ablated, while the other layers remain unablated. It may come with a small disclaimer warning. The size after conversion may differ from the original GGUF. The newly added Huihui-Qwen3.8-27B-abliterated-UD series come from unsloth/Qwen3.8-27B-GGUF. Only layers 18 to 51 have been ablated(Previously…
Qwen3.8-27B uncensored by HauhauCS 0/465 Refusals. This is the Aggressive variant: direct answers, no refusal behavior, and minimal preamble on hard prompts. Every text GGUF preserves Qwen3.8's native NextN head, and this release adds HauhauCS FastMTP: a specific acceleration sidecar qualified across the complete quant lineup at maximum native context. Vision is included through the separate BF16 projector. No changes to datasets or intended capabilities. This release preserves Qwen3.8-27B's text, reasoning, agentic, image, and video capabilities while applying the HauhauCS Aggressive uncensoring profile. Pick Aggressive when you specifically want the model to get to the answer without…
Gemma 4 E4B-IT uncensored by HauhauCS. 0/465 Refusals\ No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. These are meant to be the best lossless uncensored models out there. Stronger uncensoring — model is fully unlocked and won't refuse prompts. May occasionally append short disclaimers (baked into base model training, not refusals) but full content is always generated. For a more conservative uncensor that keeps some safety guardrails, check the Balanced variant when it's available. All quants generated with importance matrix (imatrix) for optimal quality preservation on abliterated weights. KP ("Perfect")…
You can now also fine-tune the model locally with Unsloth. - Read our Qwen3.5 fine-tuning guide here. Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance. Qwen3.5 represents a significant leap forward, integrating breakthroughs in multimodal learning, architectural efficiency, reinforcement learning scale, and global accessibility to empower developers and enterprises with unprecedented capability and efficiency. For more details, please refer to our blog post Qwen3.5. WMT24++: a harder subset of WMT24 after difficulty labeling and rebalancing; we report the averaged scores on 55 languages using XCOMET-XXL. Empty…