If you are training on data labeled by frontier models, how do you expect to exceed the performance of frontier models, other than in the cost dimension by recognizing simpler problems and routing to cheaper models?
show comments
YuechenLi
Have you compared this to using GPT-6.1 Sol instead of GPT 6 Astra + Deepseek? From my test, 6.1 Sol is a lot more token efficient than 6 Sol while being similar to Astra in performance, and I don't really find 6 Astra to be significantly better than 6/6.1 Sol for general coding as I feel 6 Astra is only noticeably better at spatial reasoning/vision compared to 6 Sol, and 6.1 Sol really closed the gap on that front.
gitowiec
Can it route to locally or LAN hosted Qwen or some other open weights model?
show comments
ajspig1
How do you handle provider variance on OpenRouter for the opensource models? Or do you use your own hosted version to mitigate this?
And for both opensource and closed source, does the router account for provider quality, or catch it when a provider degrades?
show comments
rirze
How does this choose which models to use with any arbitrary set of model providers to work from? And why is an openrouter necessary for self-hosting?
show comments
jamesforestwest
How do you define the model buckets, and what happens when a session genuinely needs a model that isn't in the bucket the HMM picked?
show comments
thefourthchime
Interesting work, and thanks for describing how your router works internally. It's definitely a fascinating subject. How would you say this compares to Cursor's auto mode?
show comments
1minusp
Does this allow for a predefined budget?
show comments
redrove
Is the model you trained available as open weights?
If you are training on data labeled by frontier models, how do you expect to exceed the performance of frontier models, other than in the cost dimension by recognizing simpler problems and routing to cheaper models?
Have you compared this to using GPT-6.1 Sol instead of GPT 6 Astra + Deepseek? From my test, 6.1 Sol is a lot more token efficient than 6 Sol while being similar to Astra in performance, and I don't really find 6 Astra to be significantly better than 6/6.1 Sol for general coding as I feel 6 Astra is only noticeably better at spatial reasoning/vision compared to 6 Sol, and 6.1 Sol really closed the gap on that front.
Can it route to locally or LAN hosted Qwen or some other open weights model?
How do you handle provider variance on OpenRouter for the opensource models? Or do you use your own hosted version to mitigate this?
And for both opensource and closed source, does the router account for provider quality, or catch it when a provider degrades?
How does this choose which models to use with any arbitrary set of model providers to work from? And why is an openrouter necessary for self-hosting?
How do you define the model buckets, and what happens when a session genuinely needs a model that isn't in the bucket the HMM picked?
Interesting work, and thanks for describing how your router works internally. It's definitely a fascinating subject. How would you say this compares to Cursor's auto mode?
Does this allow for a predefined budget?
Is the model you trained available as open weights?
AGI is here!