The State of Model Routing — NVIDIA, Cognition, OpenRouter
The video discusses choosing between large and small AI models for different tasks.
It shows that selecting the cheapest model for each individual task does not always pay off — a small model operating outside its area of competence can become very expensive by calling helper functions repeatedly. Cognition solves this by having a large model plan and delegate work to smaller models, which saved 40 percent on costs. Another finding is that it is valuable to keep the same model active throughout an entire conversation — this saves money because already processed information can be reused instead of being reprocessed.
Vibekollen prepared this summary with AI from the original publication. The content belongs to AI Engineer.
More to read
One word for AI agents in 2027
Nate Herk 4 h ago
Tests: chatbots rely on three obscure AI-built voter sites for US midterm answers; Copilot cited the error-prone Ballot Brief in 75% of tests and ChatGPT in 22%
Techmeme 5 h ago
Build an AI Code Reviewer with OpenRouter
OpenRouter 5 h ago
AI:AM: Carbon in the Arena, Mistral's Le Chonk, the Agent-Run Economy, Ferroelectric RAM + New Hacks
The Cognitive Revolution 7 h ago