GPT-5 might have topped all benchmarks, but people seem to believe that this lead won’t even last a month.
Polymarket flipped to showing Google as the top AI model at the end of August as GPT-5’s livestream was ongoing. On the poll “which company has the best AI model end of August?”, OpenAI was leading with around 80% when GPT-5’s livestream began, and Google was languishing at second place at around 20%. But as the presentation went on, the results flipped — Polymarket now said that Google had an 80% chance of being the top AI model by the end of August, while OpenAI only had less than 20% chance.
This seems to indicate that Polymarket wasn’t particularly impressed with GPT-5’s performance. People on the platform seemed to believe that while GPT-5 was the top model on most benchmarks as of 8th August, it was likely that Google would release a model before the end of the month which would go past GPT-5 in ability.
This might be down to how close GPT-5 ended up being to other models on benchmarks. On Artificial Analysis’s Intelligence Index, GPT-5 with ‘max’ reasoning scored 69. This was just one higher than the 68 Grok 4 had managed, two higher than the 67 managed by 03 and o4-mini, and four higher than the 65 scored by Gemini 2.5 Pro.
Polymarket seems to believe that Google, which had released Gemini 2.5 Pro five months ago in March, could soon release a model that could leapfrog the competition. It does seem likely that this could happen — Grok 4 was released just last month, and GPT-5 was released today. It’s only Google in the list that’s due a new release, and people seem to believe that its latest iteration will be ahead of GPT-5 on the benchmarks. It remains to be seen if that happens — Google has been releasing impressive models like Genie in recent times — but it appears that the collective intelligence of Polymarket believes that Google will again be on top within the next 23 days.