Step 3.7 Flash needs about 136.7 GB of memory at Q4_K_M (a 121.6 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 29 quants, from 0.5 GB to 212.0 GB (Q8_0).
Step 3.7 Flash – 199B sparse MoE vision‑language model
Step 3.7 Flash is a 199B‑parameter sparse Mixture‑of‑Experts model with a 196B language backbone and a 1.8B vision encoder. It supports a 262144‑token context window and can activate about 11B parameters per token. The model is licensed under apache‑2.0.
Good at
- Multimodal perception and verification
- Tool orchestration and workflow integrity
- Live engineering and coding tasks
Limits
- Requires large VRAM for high‑quant models
- No native vision support in low‑quant variants
- Limited to 262144‑token context
Best for
- High‑throughput agentic workflows
- Parsing large documents in a single pass
- Concurrent coding agents
Who it’s for. Developers building autonomous agents that combine perception, search, and reasoning. Suitable for production workloads that need fast, reliable multimodal inference.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.5 GB | 1.0 GB | Fits | Get file → | |
| IQ1_S | 43.0 GB | 48.6 GB | Too big | Get file → | |
| IQ1_M | 47.7 GB | 53.9 GB | Too big | Get file → | |
| IQ2_XXS | 55.0 GB (2 parts) | 62.1 GB | Too big | Get file → | |
| IQ2_XS | 60.9 GB (2 parts) | 68.8 GB | Too big | Get file → | |
| IQ2_S | 62.2 GB (2 parts) | 70.2 GB | Too big | Get file → | |
| IQ2_M | 68.4 GB (2 parts) | 77.1 GB | Too big | Get file → | |
| Q2_K | 71.2 GB (2 parts) | 80.3 GB | Too big | Get file → | |
| Q2_K_L | 71.7 GB (2 parts) | 80.9 GB | Too big | Get file → | |
| IQ3_XXS | 84.2 GB (3 parts) | 94.8 GB | Too big | Get file → | |
| Q3_K_S | 87.6 GB (3 parts) | 98.6 GB | Too big | Get file → | |
| IQ3_XS | 91.7 GB (3 parts) | 103.2 GB | Too big | Get file → | |
| Q3_K_M | 91.8 GB (3 parts) | 103.3 GB | Too big | Get file → | |
| Q3_K_L | 95.5 GB (3 parts) | 107.5 GB | Too big | Get file → | |
| Q3_K_XL | 96.0 GB (3 parts) | 108.0 GB | Too big | Get file → | |
| IQ3_M | 96.0 GB (3 parts) | 108.0 GB | Too big | Get file → | |
| IQ4_XS | 107.3 GB (3 parts) | 120.7 GB | Too big | Get file → | |
| IQ4_NL | 113.3 GB (3 parts) | 127.4 GB | Too big | Get file → | |
| Q4_0 | 113.6 GB (3 parts) | 127.7 GB | Too big | Get file → | |
| Q4_K_S | 117.1 GB (3 parts) | 131.6 GB | Too big | Get file → | |
| Q4_K_M | 121.6 GB (4 parts) | 136.7 GB | Too big | Get file → | |
| Q4_K_L | 122.0 GB (4 parts) | 137.2 GB | Too big | Get file → | |
| Q4_1 | 125.3 GB (4 parts) | 140.8 GB | Too big | Get file → | |
| Q5_K_S | 137.5 GB (4 parts) | 154.4 GB | Too big | Get file → | |
| Q5_K_M | 142.0 GB (4 parts) | 159.6 GB | Too big | Get file → | |
| Q5_K_L | 142.4 GB (4 parts) | 159.9 GB | Too big | Get file → | |
| Q6_K | 171.8 GB (5 parts) | 192.9 GB | Too big | Get file → | |
| Q6_K_L | 172.0 GB (5 parts) | 193.2 GB | Too big | Get file → | |
| Q8_0 | 212.0 GB (6 parts) | 237.9 GB | Too big | Get file → |
For 8‑GB VRAM, use IQ1_S (42.99 GB file, 48.6 GB memory at 8K context). For 16‑GB VRAM, IQ2_M (68.38 GB file, 77.1 GB memory) or Q2_K (71.23 GB file, 80.3 GB memory). For 24‑GB VRAM, Q3_K_M (91.81 GB file, 103.3 GB memory) or Q4_K_M (121.62 GB file, 136.7 GB memory).
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: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 1.0 GB in use).
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 121.6 GB file.
The Q8_0 file is 212.0 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 the first one in the table at 0.5 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 3.7 Flash
How much VRAM does Step 3.7 Flash need?
About 136.7 GB at Q4_K_M with an 8K context window. That is the 121.6 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?
Yes, with the smaller files: the 0.5 GB file needs about 1.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Step 3.7 Flash GGUF?
It depends on the quant: file is 0.5 GB, IQ1_S is 43.0 GB, IQ1_M is 47.7 GB, IQ2_XXS is 55.0 GB, IQ2_XS is 60.9 GB, IQ2_S is 62.2 GB and 23 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 bartowski/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 IQ1_S quant (42.99 GB file, 48.6 GB memory at 8K context) is feasible with 8 GB VRAM, but performance will be limited.
Does Step 3.7 Flash support vision tasks?
Yes, it includes a 1.8B vision encoder for native image understanding and excels in visual verification benchmarks.
What license applies to Step 3.7 Flash?
The model is released under the apache‑2.0 license.