CoolFace
Modelpublic

Vegss/Qwopus3.6-35B-A3B-v1-MTP-GGUF

sourceHugging Faceapache-2.0updated 4mo agoView on Hugging Face
0likes418downloads
Model Card

Jackrong/Qwopus3.6-35B-A3B-v1-MTP-GGUF

<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; border: 1px solid #93c5fd; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02); margin-bottom: 20px;"> <div style="background: linear-gradient(135deg, #3b82f6 0%, #1d4ed8 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"> <span>⚡</span> What is MTP (Multi-Token Prediction)? </div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.6;"> MTP (Multi-Token Prediction) is a technique introduced in the Qwen3.6 architecture that enables the model to predict multiple future tokens simultaneously. By leveraging dedicated MTP heads, this model supports <b>speculative decoding</b>, where a draft model predicts multiple tokens at once and the target model verifies them in parallel, resulting in significant inference speedups without sacrificing output quality. <br><br> This GGUF release preserves the MTP heads from <code>unsloth/Qwen3.6-35B-A3B</code>, making it compatible with mainstream inference frameworks that support MTP-based speculative decoding (such as llama.cpp and its derivatives). For optimal throughput, pair this MTP-enabled GGUF with a corresponding draft model. </div> </div>

Source model: Jackrong/Qwopus3.6-35B-A3B-v1 MTP source: unsloth/Qwen3.6-35B-A3B

Uploaded GGUF variants:

  • —Qwopus3.6-35B-A3B-v1-MTP-Q2_K.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q3_K_S.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q3_K_M.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q3_K_L.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-IQ4_XS.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q4_K_S.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q4_K_M.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q5_K_S.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q5_K_M.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q6_K.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-Q8_0.gguf
  • —Qwopus3.6-35B-A3B-v1-MTP-BF16.gguf

This release was prepared by validating or injecting Qwen MTP/nextn tensors before GGUF conversion.