3 ms·
The accuracy degrades with a larger context size, as pointed out in the article itself: > Claude offers fast inference, GPT-3.5-level accuracy, more customizat
by sourcelabs 3y ago
The accuracy degrades with a larger context size, as pointed out in the article itself:
> Claude offers fast inference, GPT-3.5-level accuracy, more customization options for large customers, and up to a 100k context window (though we’ve found accuracy degrades with the length of input).