A user on Mastodon shared their process for selecting a new LLM inference aggregator service using duck.ai. They locally ran Qwen3.8 to analyze the service's list of 128 models based on context window, cost per million tokens, and recency, assigning z-scores. Additionally, they used muse-glimmer locally to edit their Zed settings.json to integrate the chosen models into their AI provider, noting that their cost-optimization process has itself been optimized. AI
IMPACT Demonstrates a user-driven approach to optimizing LLM service selection and integration.
RANK_REASON User describes using specific tools and models for a personal workflow.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →