Theo Shares Gemini 3.8 Flash Score on DeepSWE
Theo posts screenshot of model scores on DeepSWE benchmark chart.
TLDR
Theo posted on X that Gemini 3.8 Flash scored 73.8 percent on DeepSWE v1.1 while GPT-6 Astra scored 73.3 percent. The attached chart plots five models against output token volume on a dark-mode line graph with score on the y-axis from 0.4 to 0.8. Theo expressed surprise at the result in the post. The evidence shows only the tweet content and chart description supplied in the packet.
Combined views
434.2K
4 Sources, first seen 27d ago