Unverified50% confidenceBenchmarkExact time
With tensor parallelism across four M5 Ultras, Muse Spark Open's decoding speed rose from 15.74 tokens/s under two-machine pipeline parallelism to 36.73 tokens/s, and Qwen 3.8 rose from 19.44 to 50.89 tokens/s.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
UnverifiedBased on public data for high power plus MLX or Ollama, the decoding speed of different models on M4 Max is roughly a ladder as follows: Nemotron 3 Nano about 160 tokens/s, Qwen 3.5 35B at 107 to 130 tokens/s, Gemma 4 26B about 75 tokens/s, DeepSeek V4 Flash about 27 tokens/s, and dense 27B about 22 tokens/s.79% similarUnverifiedIn short-answer scenarios, on the same M4 Mac, merely changing the power mode jumps the speed from 28.5 tokens/s to 109.6 tokens/s, a difference of 3.84 times.74% similarUnverifiedOn a 2000-token new prompt, DeepSeek V4.1 Flash at Q4 quantization starts prompt processing at approximately 360 tokens/s, dropping to approximately 307 at 64K context; token generation is stable at approximately 14 tokens/s.73% similarUnverifiedQwen3.8-27B的输出速度为55.1 tokens/秒，Ornith 1.5为90.4 tokens/秒。73% similarUnverifiedIn a fair 4-bit comparison, MLX with D-Flash 2 speculative decoding ran at 16.6 tokens/s, versus Ollama Q4's 5.7 tokens/s.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1038329API
curl https://kongchang.com/api/v1/knowledge/claims/1038329MCP
get_claim(id=1038329)