Expired80% confidenceFactTime unknown
The Qwen3.6 MTP-GGUF version pushes the inference speed of the 35B-A3B model to 220 tokens/s on a single GPU, more than 1.4× faster than the original GGUF with zero precision loss.
1
Sources
80%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedThe Qwen3.6 35B model achieved a generation speed of 86.7 tokens/s on oMLX with MTP technology, with a prompt processing speed of 1,735 tokens/s.74% similarUnverifiedQwen3.5 is 8.6× faster than the previous generation Qwen3 Max at a 32K context length, and 19× faster at a 256K context length.74% similarUnverifiedIn benchmark testing, Qwen3 27B achieved a generation speed of approximately 40-something tokens per second on 4 modified 2080 Ti cards without NVLink enabled.73% similarUnverifiedRunning the Qwen3.6 27B model on a GPU with 24 GB of VRAM, the generation speed reaches approximately 40 tokens per second without optimization, and 50–60 tokens/s after optimization.73% similarUnverifiedOn the RTX 5090, the Qwen3.6 27B model with Q4 quantization has a decoding speed of approximately 63 tokens/s without MTP, jumping to 105 tokens/s with MTP enabled.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/18049API
curl https://kongchang.com/api/v1/knowledge/claims/18049MCP
get_claim(id=18049)