Fetching from the wire…
Public story · 2026-03-18 · source-backed
Developer Dan Woods ran a frontier-class 397B MoE model on a 256GB M3 Ultra at 25+ tokens/sec using GGUF quantization and MoE layer offloading. Practical proof that frontier-tier inference is becoming achievable on consumer hardware. simonwillison.net
Each link below shares sources, entities, or timing with this story.
Anthropic partners with Apple / Shared entities / Shared topic / What happened next
Linked by a graph relationship (Anthropic partners with Apple); both cover Apple, Flash, GGUF, MoE; overlapping topics (inference, model).
Anthropic partners with Apple / Shared entities / Same source domain / Shared topic / What happened next / Tension
Linked by a graph relationship (Anthropic partners with Apple); both cover GGUF, MoE, Qwen3; reported by the same outlet (simonwillison.net).
Apple supports Claude / Shared entities / Same source domain / Shared topic / What happened next
Linked by a graph relationship (Apple supports Claude); both cover MoE, Qwen3; reported by the same outlet (simonwillison.net).
Apple criticizes OpenAI / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Apple criticizes OpenAI); both cover MoE, Qwen3; overlapping topics (consumer, hardware).
Apple partners with Google / Shared entities / What happened next
Linked by a graph relationship (Apple partners with Google); both cover Flash, GGUF, MoE; picks up the Flash thread on 2026-08-16.
Linked by a graph relationship (Apple partners with Google); both cover LLM, MoE, Qwen3; picks up the LLM thread on 2026-04-02.
Apple partners with Google / Shared entities / Shared topic / What happened next
Linked by a graph relationship (Apple partners with Google); both cover Flash, LLM; overlapping topics (flash, model).
Apple uses Gemini / Shared entities / Same source domain / What happened next
Linked by a graph relationship (Apple uses Gemini); both cover Flash, LLM; reported by the same outlet (simonwillison.net).