INFORMATIONAL Model ranking/recommendations - May 2026

[mhenke]

Findings: A community member shared an extensive May-2026 model-ranking guide for each agent role, built from public benchmarks (Artificial Analysis, LMSYS Arena, SWE-bench, Terminal-Bench, and others). It is offered as a reference, not a request for changes.

What was said:

• @FirebirdRender: posted tiered rankings (reasoning, complex, standard, simple) with top-5 closed and open-weight picks per role, plus specialist picks for Observer, Designer, and Librarian and a quick-reference table. Noted the rankings shift monthly and are approximate.

Open questions:

1. How much do these May-2026 rankings still hold as models release, and should we keep a living version somewhere?
2. For a quality-first setup, do the recommended Oracle, Designer, and Librarian picks match what people actually run day to day?
3. Should the plugin docs link to or vendor a curated model-ranking reference, or leave it to community discussions?

Original issue: #498

[mhenke]

Consolidated into #751