ST

Step-3.7-Flash

StepGGUF by unslothUpdated 2026-05-30
Download on Hugging Face
Parameters197B
Context256K
Layers45
Architecturestep35
Smallest quant56.8 GB
Quick answer

Step-3.7-Flash needs about 137.2 GB of memory at Q4_K_M (a 122.1 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 22 quants, from 56.8 GB (IQ1_M) to 394.0 GB (BF16).

Step‑3.7‑Flash: 198 B‑parameter vision‑language MoE model

Step‑3.7‑Flash is a 198 B‑parameter sparse Mixture‑of‑Experts vision‑language model. It combines a 196 B language backbone with a 1.8 B vision encoder and supports a 256k context window. The model activates about 11 B parameters per token and can process up to 400 tokens per second.

Good at

  • High‑throughput multimodal inference
  • Large‑context reasoning
  • Agentic workflow orchestration

Limits

  • Requires large VRAM for full‑precision quants
  • No native vision support in lower‑rank quants
  • License is Apache‑2.0

Best for

  • Parsing large documents in one pass
  • Running multi‑step search loops
  • Concurrent coding agents

Who it’s for. Developers building agentic systems that combine perception, search, and reasoning. Suitable for production workloads needing high throughput and large context.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_M56.8 GB (3 parts) 64.1 GB
Too big Get file →
IQ2_XXS61.7 GB (3 parts) 69.6 GB
Too big Get file →
IQ2_M61.8 GB (3 parts) 69.7 GB
Too big Get file →
Q2_K_XL65.8 GB (3 parts) 74.2 GB
Too big Get file →
IQ3_XXS73.6 GB (3 parts) 82.9 GB
Too big Get file →
IQ3_S80.0 GB (3 parts) 90.1 GB
Too big Get file →
Q3_K_M89.3 GB (3 parts) 100.5 GB
Too big Get file →
Q3_K_XL89.4 GB (3 parts) 100.6 GB
Too big Get file →
IQ4_XS95.3 GB (3 parts) 107.3 GB
Too big Get file →
IQ4_NL97.3 GB (3 parts) 109.5 GB
Too big Get file →
Q4_K_S114.2 GB (4 parts) 128.4 GB
Too big Get file →
MXFP4_MOE118.3 GB (4 parts) 132.9 GB
Too big Get file →
Q4_K_M122.1 GB (4 parts) 137.2 GB
Too big Get file →
Q4_K_XL122.2 GB (4 parts) 137.4 GB
Too big Get file →
Q5_K_S138.1 GB (4 parts) 155.1 GB
Too big Get file →
Q5_K_M146.5 GB (4 parts) 164.6 GB
Too big Get file →
Q5_K_XL146.5 GB (4 parts) 164.6 GB
Too big Get file →
Q6_K163.3 GB (5 parts) 183.4 GB
Too big Get file →
Q6_K_XL178.7 GB (5 parts) 200.6 GB
Too big Get file →
Q8_0209.4 GB (6 parts) 235.0 GB
Too big Get file →
Q8_K_XL209.6 GB (6 parts) 235.3 GB
Too big Get file →
BF16394.0 GB (8 parts) 441.8 GB
Too big Get file →

For 8 GB VRAM, use IQ1_M (56.75 GB file, ~64 GB memory at 8k context). For 16 GB VRAM, Q2_K_XL (65.79 GB file, ~74 GB memory). For 24 GB VRAM, Q4_K_M (122.09 GB file, ~137 GB memory). Memory estimates are approximate.

What GPU do you need for Step-3.7-Flash?

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-3.7-Flash quant should you choose?

Q4_K_M is the usual starting point for Step-3.7-Flash: it balances file size and quality well. Here it is a 122.1 GB file.

The Q8_0 file is 209.4 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is IQ1_M at 56.8 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturestep35
Parameters197B (approximate, mixture of experts)
Layers45
Trained context262,144 tokens

Source

Quantized byunsloth
FamilyStep
Licenseapache-2.0
Downloads (HF)11,521
Last updated2026-05-30

Frequently asked questions about Step-3.7-Flash

How much VRAM does Step-3.7-Flash need?

About 137.2 GB at Q4_K_M with an 8K context window. That is the 122.1 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-3.7-Flash on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 64.1 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Step-3.7-Flash GGUF?

It depends on the quant: IQ1_M is 56.8 GB, IQ2_XXS is 61.7 GB, IQ2_M is 61.8 GB, Q2_K_XL is 65.8 GB, IQ3_XXS is 73.6 GB, IQ3_S is 80.0 GB and 16 more.

What context length does Step-3.7-Flash support?

The GGUF header lists a trained context length of 262,144 tokens (256K).

Where can I download Step-3.7-Flash GGUF?

The files are on Hugging Face in the unsloth/Step-3.7-Flash-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-3.7-Flash use?

Hugging Face lists the license as apache-2.0. 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‑3.7‑Flash on 8 GB VRAM?

Only the lowest‑rank quants such as IQ1_M can fit, but memory usage will still exceed 8 GB at 8k context. Consider a larger GPU or a smaller quant.

Does the model support vision tasks?

Yes, the model includes a 1.8 B vision encoder and is capable of native image understanding.

What license applies to this model?

The model is released under the Apache‑2.0 license.

Other GGUFs of Step-3.7-Flash

Similar models

Browse all →

More from unsloth

All by unsloth →

More Step models

All Step →

Other large models

All large models →