Step-5-Preview-BF16 needs about 308.7 GB of memory at its main file (a 275.2 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 5 quants, from 248.8 GB to 333.0 GB.
Step-5-Preview-BF16 – 601 billion‑parameter vision‑enabled model
Step-5-Preview-BF16 is a 601 billion‑parameter model with vision capabilities. It is trained on a context of 1 048 576 tokens and is a mixture of experts. The model is available in three GGUF shards with sizes 231.32 GB, 275.17 GB, and 332.97 GB.
Good at
- Vision tasks
- Large context handling
- Mixture‑of‑experts inference
Limits
- No language metadata provided
- License type is other
- Requires large VRAM for full model
Best for
- Vision‑heavy inference
- Long‑form content generation
- Expert‑level image analysis
Who it’s for. Developers needing high‑capacity vision models for research or production. Suitable for environments with sufficient GPU memory to load the full 601 billion‑parameter model.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 248.8 GB (27 parts) | 279.1 GB | Too big | Get file → | |
| — | 275.2 GB (30 parts) | 308.7 GB | Too big | Get file → | |
| — | 275.2 GB (30 parts) | 308.7 GB | Too big | Get file → | |
| — | 333.0 GB (36 parts) | 373.4 GB | Too big | Get file → | |
| — | 333.0 GB (36 parts) | 373.4 GB | Too big | Get file → |
For 8 GB VRAM, use the 231.32 GB shard; it estimates 259.6 GB memory at 8 k context. For 12 GB VRAM, the 275.17 GB shard estimates 308.7 GB memory. For 16 GB VRAM, the 332.97 GB shard estimates 373.4 GB memory. These estimates are approximate.
What GPU do you need for Step-5-Preview-BF16?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which Step-5-Preview-BF16 quant should you choose?
If memory is tight, the smallest file is the first one in the table at 248.8 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Step-5-Preview-BF16
How much VRAM does Step-5-Preview-BF16 need?
About 308.7 GB at its main file with an 8K context window. That is the 275.2 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Step-5-Preview-BF16 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 279.1 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Step-5-Preview-BF16 GGUF?
It depends on the quant: file is 248.8 GB, file is 275.2 GB, file is 275.2 GB, file is 333.0 GB, file is 333.0 GB.
What context length does Step-5-Preview-BF16 support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K).
Where can I download Step-5-Preview-BF16 GGUF?
The files are on Hugging Face in the L-Alchemyst/Step-5-Preview-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Step-5-Preview-BF16 use?
Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Step-5-Preview-BF16 on 8 GB VRAM?
The 231.32 GB shard estimates 259.6 GB memory at 8 k context, so 8 GB VRAM is insufficient.
What is the context length of Step-5-Preview-BF16?
The model is trained on 1 048 576 tokens.
Does Step-5-Preview-BF16 support text generation?
The model has vision capabilities but no language metadata is provided.