Step 3.5 Flash REAP 121B A11B i1 needs about 82.1 GB of memory at Q4_K_M (a 72.8 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 24 quants, from 0.3 GB to 99.3 GB (Q6_K).
Step-3.5-Flash-REAP-121B-A11B: 121B sparse MoE model with 11B active experts
Step-3.5-Flash-REAP-121B-A11B is a 121B parameter sparse Mixture-of-Experts causal language model. It uses REAP pruning to keep 11B active parameters per token and supports a 262,144 token context. The model is compatible with vanilla vLLM and requires no source changes.
Good at
- Large context handling
- Sparse expert inference
- Tool calling
Limits
- Requires high VRAM for large quants
- No vision support
- License is other
Best for
- Local deployment with limited GPU memory
- Academic research on sparse models
- Code generation tasks
Who it’s for. Developers and researchers who need a memory‑efficient version of Step-3.5-Flash with near‑lossless performance. The model is suitable for environments where GPU memory is constrained but large context and expert routing are required.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.3 GB | 0.8 GB | Fits | Get file → | |
| IQ1_S | 24.4 GB | 27.8 GB | Too big | Get file → | |
| IQ1_M | 27.1 GB | 30.9 GB | Too big | Get file → | |
| IQ2_XXS | 31.7 GB | 36.0 GB | Too big | Get file → | |
| IQ2_XS | 35.4 GB | 40.2 GB | Too big | Get file → | |
| IQ2_S | 35.9 GB | 40.7 GB | Too big | Get file → | |
| IQ2_M | 39.5 GB | 44.7 GB | Too big | Get file → | |
| Q2_K_S | 40.6 GB | 45.9 GB | Too big | Get file → | |
| Q2_K | 44.2 GB | 50.0 GB | Too big | Get file → | |
| IQ3_XXS | 46.6 GB | 52.7 GB | Too big | Get file → | |
| IQ3_XS | 49.3 GB | 55.7 GB | Too big | Get file → | |
| Q3_K_S | 52.3 GB | 59.1 GB | Too big | Get file → | |
| IQ3_S | 52.3 GB | 59.1 GB | Too big | Get file → | |
| IQ3_M | 52.9 GB | 59.7 GB | Too big | Get file → | |
| Q3_K_M | 57.7 GB | 65.1 GB | Too big | Get file → | |
| Q3_K_L | 62.8 GB | 70.8 GB | Too big | Get file → | |
| IQ4_XS | 64.6 GB | 72.8 GB | Too big | Get file → | |
| Q4_0 | 68.4 GB | 77.1 GB | Too big | Get file → | |
| Q4_K_S | 68.5 GB | 77.3 GB | Too big | Get file → | |
| Q4_K_M | 72.8 GB | 82.1 GB | Too big | Get file → | |
| Q4_1 | 75.8 GB | 85.4 GB | Too big | Get file → | |
| Q5_K_S | 83.3 GB | 93.8 GB | Too big | Get file → | |
| Q5_K_M | 85.7 GB | 96.5 GB | Too big | Get file → | |
| Q6_K | 99.3 GB | 111.8 GB | Too big | Get file → |
For 8 GB VRAM, use the IQ1_S quant (24.41 GB file, 27.8 GB memory at 8 k context). For 16 GB VRAM, IQ2_XS (35.41 GB file, 40.2 GB memory) or Q2_K (44.23 GB file, 50 GB memory) are suitable. For 32 GB VRAM, Q4_K_M (72.82 GB file, 82.1 GB memory) or Q5_K_M (85.67 GB file, 96.5 GB memory) provide higher quality.
What GPU do you need for Step 3.5 Flash REAP 121B A11B i1?
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 (0.3 GB file, about 0.8 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 0.8 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 0.8 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 0.8 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 0.8 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 0.8 GB in use).
- 32 GB VRAM: the best quant that fits is IQ1_M (27.1 GB file, about 30.9 GB in use).
- 48 GB VRAM: the best quant that fits is Q2_K_S (40.6 GB file, about 45.9 GB in use).
Which Step 3.5 Flash REAP 121B A11B i1 quant should you choose?
Q4_K_M is the usual starting point for Step 3.5 Flash REAP 121B A11B i1: it balances file size and quality well. Here it is a 72.8 GB file.
If memory is tight, the smallest file is the first one in the table at 0.3 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 3.5 Flash REAP 121B A11B i1
How much VRAM does Step 3.5 Flash REAP 121B A11B i1 need?
About 82.1 GB at Q4_K_M with an 8K context window. That is the 72.8 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 3.5 Flash REAP 121B A11B i1 on an 8 GB GPU?
Yes, with the smaller files: the 0.3 GB file needs about 0.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Step 3.5 Flash REAP 121B A11B i1 GGUF?
It depends on the quant: file is 0.3 GB, IQ1_S is 24.4 GB, IQ1_M is 27.1 GB, IQ2_XXS is 31.7 GB, IQ2_XS is 35.4 GB, IQ2_S is 35.9 GB and 18 more.
What context length does Step 3.5 Flash REAP 121B A11B i1 support?
The GGUF header lists a trained context length of 262,144 tokens (256K).
Where can I download Step 3.5 Flash REAP 121B A11B i1 GGUF?
The files are on Hugging Face in the mradermacher/Step-3.5-Flash-REAP-121B-A11B-i1-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 3.5 Flash REAP 121B A11B i1 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-3.5-Flash-REAP-121B-A11B on 8 GB VRAM?
The IQ1_S quant uses 27.8 GB memory at 8 k context, so 8 GB VRAM is insufficient. You would need a larger VRAM or a smaller quant.
Does this model support vision tasks?
No, the model has no vision capability.
What is the context length of this model?
The model supports 262,144 tokens of context.