Models

Models is the canonical Hax hub for this subject. It is not a tag archive: it explains the whole topic first, then routes readers to narrower cluster pages for the exact job they are trying to finish. Start here when you need the map: what the category means, which parts are mature enough to use, where local AI still breaks, and which Hax tests support the recommendation.

The cluster is organized so authority concentrates on this page instead of being split across near-duplicate articles. Broad questions stay on the pillar; specific intents move to cluster pages such as measured benchmarks, setup guides, comparisons, operational retrospectives, and concept explainers. Each cluster page links back to this pillar through its topic breadcrumb and kicker, while this pillar links out with descriptive anchors. That bidirectional structure helps crawlers and readers see one coherent body of work rather than a loose feed.

For Models, the useful reading order is simple: read this summary to get the decision frame, scan the section that matches your intent, then open the deepest article for measurements, source links, and caveats. Current cluster coverage includes 63 published Hax posts including Fish Speech Quality Drop, Confirmed by Measured Benchmarks, How We Keep Benchmark Numbers at Zero Fabrication, The AI Stack One Team Runs, By the Numbers, Our LLM serving: throughput vs latency cross-check. Thin overlap is treated as consolidation fuel: when several posts chase the same question, the stronger URL acts as the authority center and supporting pages point back into the same canonical path.