Step 3.7 Flash ARA Heretic needs about 118.1 GB of memory at IQ4_XS (a 105.0 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 2 quants, from 105.0 GB (IQ4_XS) to 209.4 GB (Q8_0).
Step 3.7 Flash ARA Heretic – 198B sparse MoE vision‑language model
Step 3.7 Flash ARA Heretic is a 198B‑parameter sparse Mixture‑of‑Experts model. It combines a 196B‑parameter language backbone with a 1.8B‑parameter vision encoder. The model supports a 256k context window and delivers up to 400 tokens per second.
Good at
- Vision‑language reasoning
- High‑throughput agentic workflows
- Large context handling
Limits
- Requires 118.1GB VRAM at 8k context for IQ4_XS
- Requires 235GB VRAM at 8k context for Q8_0
- No native vision support in non‑vision quantizations
Best for
- Parsing large financial reports
- Multi‑step search loops with verification
- Concurrent coding agents in pipelines
Who it’s for. Developers building production‑grade agentic systems that combine perception, search, and reasoning. It is suited for tasks that demand high throughput and large context windows.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ4_XS | 105.0 GB (3 parts) | 118.1 GB | Too big | Get file → | |
| Q8_0 | 209.4 GB (5 parts) | 235.0 GB | Too big | Get file → |
For a 8GB VRAM budget, the model cannot run. If you have a 120GB VRAM system, IQ4_XS is the smallest quantization at 104.99GB file size and 118.1GB memory at 8k context. For larger GPUs, Q8_0 offers higher fidelity with a 209.42GB file size and 235GB memory at 8k context.
What GPU do you need for Step 3.7 Flash ARA Heretic?
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 ARA Heretic quant should you choose?
IQ4_XS (105.0 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Step 3.7 Flash ARA Heretic
How much VRAM does Step 3.7 Flash ARA Heretic need?
About 118.1 GB at IQ4_XS with an 8K context window. That is the 105.0 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 ARA Heretic on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 118.1 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Step 3.7 Flash ARA Heretic GGUF?
It depends on the quant: IQ4_XS is 105.0 GB, Q8_0 is 209.4 GB.
What context length does Step 3.7 Flash ARA Heretic support?
The GGUF header lists a trained context length of 262,144 tokens (256K).
Where can I download Step 3.7 Flash ARA Heretic GGUF?
The files are on Hugging Face in the PeterCxy/Step-3.7-Flash-ARA-Heretic-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 ARA Heretic 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 8GB VRAM?
No, the smallest quantization requires 118.1GB VRAM at 8k context.
What is the maximum context length?
The model supports a 256k context window.
Is the model licensed under Apache 2.0?
Yes, the model is released under the apache-2.0 license.