cerebras/Qwen3-Coder-REAP-363B-A35B needs about 221.9 GB of memory at Q4_K_M (a 219.3 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 25 quants, from 101.3 GB (TQ1_0) to 726.3 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.3 GB.
Qwen3-Coder-REAP-363B-A35B: 25% lighter compressed Qwen3-Coder with near‑lossless coding performance
Qwen3-Coder-REAP-363B-A35B is a memory‑efficient compressed variant of Qwen3-Coder-480B-A35B-Instruct. It uses REAP pruning to reduce parameters from 480B to 363B while keeping 35B active experts per token. The model supports 262,144 token context and is licensed under Apache 2.0.
Good at
- Code generation
- Agentic coding
- Function calling
Limits
- Large context length requires high VRAM
- No vision support
- Requires vLLM v0.11.0 or newer
Best for
- Local code assistant
- Repository‑scale code analysis
- Function‑call integration
Who it’s for. Developers and researchers needing a high‑capacity coding model that fits within tighter GPU memory budgets. It is suitable for local deployments and academic experiments where full 480B is impractical.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 101.3 GB | 103.9 GB | Too big | Get file → | |
| IQ1_S | 104.4 GB (3 parts) | 107.0 GB | Too big | Get file → | |
| IQ1_M | 115.8 GB (3 parts) | 118.4 GB | Too big | Get file → | |
| Q2_K | 132.2 GB (3 parts) | 134.8 GB | Too big | Get file → | |
| Q2_K_L | 132.5 GB (3 parts) | 135.0 GB | Too big | Get file → | |
| Q2_K_XL | 138.6 GB (3 parts) | 141.1 GB | Too big | Get file → | |
| IQ3_XXS | 154.8 GB (4 parts) | 157.4 GB | Too big | Get file → | |
| Q3_K_S | 156.6 GB (4 parts) | 159.1 GB | Too big | Get file → | |
| Q3_K_XL | 162.3 GB (4 parts) | 164.9 GB | Too big | Get file → | |
| Q3_K_M | 173.5 GB (4 parts) | 176.1 GB | Too big | Get file → | |
| IQ4_XS | 197.1 GB (4 parts) | 199.7 GB | Too big | Get file → | |
| IQ4_NL | 204.7 GB (5 parts) | 207.3 GB | Too big | Get file → | |
| Q4_0 | 205.5 GB (5 parts) | 208.1 GB | Too big | Get file → | |
| Q4_K_S | 206.3 GB (5 parts) | 208.9 GB | Too big | Get file → | |
| Q4_K_XL | 209.0 GB (5 parts) | 211.6 GB | Too big | Get file → | |
| Q4_K_M | 219.3 GB (5 parts) | 221.9 GB | Too big | Get file → | |
| Q4_1 | 227.3 GB (5 parts) | 229.9 GB | Too big | Get file → | |
| Q5_K_S | 249.9 GB (6 parts) | 252.5 GB | Too big | Get file → | |
| Q5_K_M | 257.5 GB (6 parts) | 260.1 GB | Too big | Get file → | |
| Q5_K_XL | 260.2 GB (6 parts) | 262.8 GB | Too big | Get file → | |
| Q6_K | 298.0 GB (7 parts) | 300.6 GB | Too big | Get file → | |
| Q6_K_XL | 312.6 GB (7 parts) | 315.2 GB | Too big | Get file → | |
| Q8_0 | 386.0 GB (8 parts) | 388.5 GB | Too big | Get file → | |
| Q8_K_XL | 425.8 GB (9 parts) | 428.3 GB | Too big | Get file → | |
| BF16 | 726.3 GB (16 parts) | 728.9 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_S (208.9 GB memory estimate) or Q4_K_M (221.9 GB). For 16 GB VRAM, Q4_K_XL (211.6 GB) or Q4_K_M (221.9 GB) are suitable. Larger VRAM allows Q5_K_S (252.5 GB) or Q5_K_M (260.1 GB).
What GPU do you need for cerebras/Qwen3-Coder-REAP-363B-A35B?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) 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 cerebras/Qwen3-Coder-REAP-363B-A35B quant should you choose?
Q4_K_M is the usual starting point for cerebras/Qwen3-Coder-REAP-363B-A35B: it balances file size and quality well. Here it is a 219.3 GB file.
The Q8_0 file is 386.0 GB, about 1.8 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 TQ1_0 at 101.3 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about cerebras/Qwen3-Coder-REAP-363B-A35B
How much VRAM does cerebras/Qwen3-Coder-REAP-363B-A35B need?
About 221.9 GB at Q4_K_M with an 8K context window. That is the 219.3 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 cerebras/Qwen3-Coder-REAP-363B-A35B on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 103.9 GB, so you would need a larger GPU or CPU offloading.
What is the file size of cerebras/Qwen3-Coder-REAP-363B-A35B GGUF?
It depends on the quant: TQ1_0 is 101.3 GB, IQ1_S is 104.4 GB, IQ1_M is 115.8 GB, Q2_K is 132.2 GB, Q2_K_L is 132.5 GB, Q2_K_XL is 138.6 GB and 19 more.
What context length does cerebras/Qwen3-Coder-REAP-363B-A35B support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.1 GB at 8K and 8.3 GB at 32K.
Where can I download cerebras/Qwen3-Coder-REAP-363B-A35B GGUF?
The files are on Hugging Face in the unsloth/Qwen3-Coder-REAP-363B-A35B-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 cerebras/Qwen3-Coder-REAP-363B-A35B 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 Qwen3-Coder-REAP-363B-A35B on 8 GB VRAM?
The smallest quant for 8 GB VRAM is Q4_K_S, which estimates 208.9 GB memory at 8k context. It is unlikely to fit on 8 GB VRAM without further optimization.
Does this model support image input?
No, the model does not include vision capabilities.
What is the maximum context length?
The native context length is 262,144 tokens, extendable to 1M with YaRN.