@notjazii Seen this post around a lot but I'll still pitch regardless
I'm building healthcare startup that aims to give overburdened hospitals and imaging centres in Africa access to radiologist for interpreting scans. It's called Nurrad
With all the performance jumps we see here from Astra, which even make Fable 5.1 struggle to compete a bit, Muse Spark 1.3 is higher on the Artificial Analysis Index. Benchmaxxing at its peak
https://t.co/au6pMLAUzB
@notjazii Just goes to show how gamified and benchmaxxed everything is. I think OpenAI actually built the model to be genuinely good from the ground up with the new pretrain. The hallucination rates are also lower now with Astra.
@bridgemindai Anthropic did say in their last technical report that the model they were testing internally did outperform fable 5 but might not be as big of a leap as you would think.