A model router for Claude Code, benchmarked: 82% cheaper, 2 tests lost

Описание и применение

Роутер Rebel Router снижает стоимость запросов в Claude Code до 82 % от стоимости Opus 5.5, улучшая латентность и сохраняя почти такой же уровень качества.

Задача

Users pick a single model for all Claude Code requests, paying high prices for simple tasks that rarely benefit from expensive models.

Результат

Router reduced average cost per task from $0.96 to $0.17 (82 % cheaper), lowered mean latency from 31.7 s to 19.5 s, and maintained 75 % pass rate versus 92 % for Opus 5.5, with three failures due to token limits.

Рабочий процесс

• Clone rebel‑router repo
• Install dependencies and link CLI
• Configure .env with provider keys
• Start router on 127.0.0.1:8787
• Initialize router for Claude Code or OpenCode
• Set per‑project budgets and rules in rebel.yaml
• Run tasks through router
• Collect ledger and dashboard data
• Benchmark against Opus 5.5

Доказательства

Rebel Router runs on your machine (127.0.0.1:8787) between your coding agent and the model providers

Источник

Router vs. Opus for everything: Cost: 0.17 USD with the router choosing, against 0.96 USD with Opus 5.5 on every task: 82 % less

Источник

Router vs. Opus for everything: Quality: 9 of 12 passed (75 %) against 11 of 12 (92 %). The three misses were two unit-test tasks and one QA task the router sent to Claude Haiku 5.5, whose answers ran past the 8,000-token limit.

Источник