lovedheart/Qwen3-Coder-Next-REAP-48B-A3B-GGUF

• Qwen/Qwen3-Coder-Next

• text-generation-inference

!Qwen3-coder-next-reap

Qwen3-Coder-Next-REAP-48B-A3B has the following specifications:

• Type: Causal Language Models
• Number of Parameters: 48B in total and 3B activated
• Hidden Dimension: 2048
• Number of Layers: 48
• Hybrid Layout: 12 (3 (Gated DeltaNet -> MoE) -> 1 * (Gated Attention -> MoE))
• Gated Attention:
• Number of Attention Heads: 16 for Q and 2 for KV
• Head Dimension: 256
• Rotary Position Embedding Dimension: 64
• Gated DeltaNet:

• Mixture of Experts:
• Number of Experts: 308 (uniformly pruned from 512)
• Number of Activated Experts: 10
• Number of Shared Experts: 1
• Context Length: 262,144 natively
• Compression Method: REAP (Router-weighted Expert Activation Pruning)
• Compression Ratio: 40% expert pruning

Test video 1 (agentic task) @Q4KXL : https://www.bilibili.com/video/BV1f8cNzcEHV/
Prompt: please clone the repository https://github.com/ggml-org/llama.cpp in /home/lovedheart/llama_ and review the PR 19435.

Test video 2 -> fastllm (int8 quantization) approx. Q8_0 in GGUF : https://www.bilibili.com/video/BV1hwFJzXEVP/
Prompt: Create a cosmic nebula background using Three.js with the following requirements: a deep black space background with twinkling white stars; 2–3 large semi-transparent purple/pink nebula clouds with a smoky texture; slow rotation animation; optimized for white text display. Implementation details: 1. Starfield: 5000 white particles randomly distributed with subtle twinkling; 2. Nebula: 2–3 large purple particle clusters using additive blending mode; 3. Colors: #8B5CF6, #C084FC, #F472B6 (purple to pink gradient); 4. Animation: overall rotation.y += 0.001, stars' opacity flickering; 5. Setup: WebGLRenderer with alpha:true and black background.