revenant.

The Arena

The same autopsy, three open-weight models, side by side. Which ghost writes the best eulogy? Pick a package and vote — every vote is logged for the receipts.

Precomputed bouts (snapshot 2026-10-30):
snapshot from 2026-10-30
gemma-aistudio1420 ms
model: gemma-4-31b-it
“A cornerstone of the Node.js ecosystem for over fifteen years, it paved the way for modern HTTP clients. Its legacy endures in the millions of applications it helped bring to life.”

classified cause: deprecated official

gemma-aistudio980 ms
model: gemma-4-26b-a4b-it
“A foundational pillar of the Node.js ecosystem that carried the weight of the web for over a decade. Its legacy remains woven into the fabric of countless modern applications.”

classified cause: deprecated official

fallback-rules14 ms
model: epitaph-heuristic-v1
“request stepped aside with a note pinned to the door. It told you where to go next — that is a kind of care.”

classified cause: deprecated official

Showing deterministic & evidence-grounded autopsies generated from the real Revenant pipeline. When live open-weight LLM gateways (Gemma, Backboard, or local Ollama) are enabled, all three models run live side by side.

How to enable additional live open-weight LLM providers

Configure GOOGLE_API_KEY (Gemma 3 27B), BACKBOARD_API_KEY (Qwen3 / Llama-3.3 via Backboard), or OLLAMA_BASE_URL (local Ollama) in your environment and set LLM_ARENA_PROVIDERS in .env.local.