Together Link: Free CLI to Swap Models in Claude

Why this matters

For developers already living inside Claude Code or Codex, Together Link removes the usual excuse for sticking with a single closed model: the switching cost. One install command, one API key, and the harness you already know how to use stays put while the model underneath becomes negotiable. That matters for founders watching token bills climb, since routing everyday tasks to GLM 5.3 Flash or DeepSeek V4.1 Flash instead of a frontier model is a direct line-item cut, not a hypothetical one. For researchers, it's a cheap way to A/B test open models like Kimi K3 against proprietary ones without rebuilding tooling each time.

The beta label is worth taking seriously, though. Commands, routing, and the model lineup are explicitly still in flux, so anyone wiring this into a production workflow today should expect breakage, not stability. Together AI has an obvious incentive here too: every token routed through Link is a token billed on their infrastructure. Worth watching whether the model list broadens, and whether Anthropic, OpenAI, or other harness makers respond kindly to a free CLI quietly rerouting traffic off their own models.