
Introducing Marathon: adaptive inference infrastructure for long-running agents, powered by @GoKiteAI. Run your agents at significantly lower cost, without sacrificing model quality. How it works: Install our one-line plugin for Codex or Claude Code, then choose a completion window per request: now, soon, later, or anytime. The more you’re willing to wait, the more you save, up to ~65% off. Launching with support for the top five open-weight models, from @Kimi_Moonshot, @Zai_org, @deepseek_ai, @Alibaba_Qwen, and @nvidia. Try Marathon today: marathon.build/




