Category Benchmarks

Benchmarks

Qwen Image 2.1: M3 Ultra vs M5 Ultra vs DGX Spark

Qwen Image 2.1 warm median render times in seconds per 1024-square image: M5 Ultra 34.826 base and 7.551 Turbo; DGX Spark 38.627 and 8.373; M3 Ultra 155.503 and 32.483.

Turbo took about eight seconds on both the M5 Ultra and Spark. Here’s what we noticed after trying the same image prompts on all three machines.

ikeo1
October 10, 2026

Benchmarks

Local LLM vs DeepSeek Cloud in a real Hermes agent job

Two cyan threads from a Mac and one silver thread from a cloud reach the same plugin list; one local thread ends in a check, the other in a mismatched tile

One read-only WordPress inventory job through the same Hermes agent, run on local Qwen and Muse on a Mac Studio and on DeepSeek Cloud. What each model did, what it got wrong, and what that says about routing.

ikeo1
September 3, 2026

Benchmarks

Five AI Models, One Coding Job: Where the Local Mac Fell Behind

Four silver tracks from a cloud reach the finish; one cyan track from a Mac stops at a barrier

Five models, one tightly bounded agentic coding job. Four cloud models finished in under 40 seconds; the local 27B on a Mac Studio did not finish. The method and every setting are attached.

ikeo1
September 1, 2026