
microsoft/Magma-8B
Image-Text-to-Text
•
Updated
•
12.3k
•
331
The training cost of Grok 3 is much higher than that of DeepSeek-R1, but the performance gap is not that big. I am not optimistic about Grok 3.
I see your point and agree. The gap in performance doesn’t seem to justify the massive cost. I have my doubts about Grok 3 as well.