Today, we announced the next step in the Project HydraFusion vision: local models and intelligent routing for GitHub Copilot on Windows. @cinnamon_msft shows what this can mean for your everyday coding workflows.

Oct 8, 2026 · 12:19 AM UTC

20
43
5
306
21,132
RelevantRecentLikes
Local models + intelligent routing is actually a big deal — faster, cheaper, and more private when the cloud isn’t needed. But “Hybrid Intelligence” instead of Super Intelligence… 😂 Welcome to the new era of terminology innovation. 🚀
1
1
79
welcome to the new era!
1
38
Hybrid routing between on-device SLMs and cloud models sounds beautiful in a keynote, but Windows workstations are already gasping for air under enterprise workloads. If GitHub Copilot starts fighting with local Docker containers, heavy JetBrains/VS Code instances, and background compilation tasks for remaining VRAM, developer velocity will tank due to local hardware bottlenecks. Unless the HydraFusion routing engine has a zero-overhead prediction layer to determine task complexity before touching the local runtime, the constant context-switching between local and cloud inference is going to introduce more jitter than pure network latency. Are we optimizing for cloud bill reduction at the expense of our developers' hardware thermal throttling?
1
353
It’s overhyped. I used the VS Code release version, spent my full Copilot Max credit on one task, and still couldn’t finish it after hours of the worst experience, with zero results.
123
Is selecting the local model available yet? Please add as an option.
211
routing is a good idea if it's invisible. the second i have to pick a model myself i'm out
185
Local plus routed is the right shape - a small local model for autocomplete, remote only for the hard steps. That split cuts the per-seat token bill more than any pricing negotiation does.
8
Load more replies