Free the Models: Harness Design at the Frontier
Replit stopped routing work for the model and let the model decide. Here are the numbers.

About Free the Models: Harness Design at the Frontier
Replit argues that model routers have a built-in ceiling: the router is always less capable than the model it is choosing for. So Replit Agent now lets the frontier model decide how hard to think, when to hand work off, and which specialist subagent (code, UI design, slides, writing) takes each step.
Why it matters: on DeepSWE, Replit Agent scored 72% at $2.11 a task against 74% at $4.43 for Astra alone; on Terminal-Bench, 49% at $2.53 against 60% at $5.86. Models also use the same freedom differently: GPT-6 Astra reused briefed workers 42% of the time, while Fable models rarely used general workers.
Read it if you are building an agent and are about to hard-code which model or step handles what.
The facts
- Type
- Guides
- Tags
- AIDeveloper ToolsEducation
- Listed
- October 6, 2026
More Guides
See all →- DEV's Agent Factory: 475 PRs, 70ms to 16msHow DEV merged 475 PRs in 30 days with an agent factory and cut Rails time to 16ms.
- Don't Write Markdown Files for Your Coding Agent. Write ScriptsRules in AGENTS.md are suggestions. A script with a pass or fail result is a constraint.
- An Agent Said It Deployed. The Containers Were a Day OldFour claims an agent makes about a deploy, and the command that proves each one.
- Jev + Opus: Frontier-Accuracy LLM Judging at 4x Lower CostHightouch's test: a Jev plus Opus hybrid judge matched frontier accuracy at over 4x lower cost.
- Agent Skills Drift: A Feedback Loop for Keeping One TrueAgent skills go stale in six ways while their checks stay green. A step-by-step loop to catch it.