Curated reference solutions and solution notes for Artificial Analysis Intelligence Index and additional evaluations, plus cybersecurity/agent benchmarks (ExploitGym, CVE-Bench, and peers) and broader agent evals (GAIA, WebArena).

Pi extension — shared model data layer (models.dev + BenchLM), cached at ~/.cache/pi

Consensus-based multi-agent debate as a native opencode plugin: heterogeneous debater sessions argue a question over annealed rounds; a coordinator judges convergence and synthesizes an answer; optimizer-analogy steering strategies and a transformers.js R

Pi extension — shared model data layer (models.dev + BenchLM), cached at ~/.cache/pi

Eval charts for Svelte — animated, themeable bar charts for showing model evaluations.

Stop waiting for Opus on every grep — routes 80% of tasks to faster models with zero quality loss. One command, zero dependencies.

MCP server for Artificial Analysis API - get LLM model pricing, speed, and benchmarks

vibeOS backend core: API server, MCP server, web dashboard, and API client

Conversation-aware model routing for OpenCode using Cast AI's Kimchi models

High-performance LLM inference in Node.js