9h ago

Cameron R. Jones and Benjamin K. Bergen report in PNAS that GPT-4.5 reached a 73 percent human judgment rate in five-minute three-party Turing tests

Llama-3.1-405B hit 56 percent while older models also cleared 50 percent.

——0——
Original post
Tuhin Chakrabarty#1050@TUHINCHAKROPCameron JonesCameron Jones@CAMROBJONES

Really excited that this is out in @PNASNews! We find that 2 LLMs (GPT-4.5 and LLama-3.1-405B) pass a 5 minute Turing test. As an update to our preprint we also find that GPT-5 and LLaMa pass a 15 minute test! 🧵

9:45 AM · May 19, 2026 View on X
POSTEric Topol#735Eric Topol@ERICTOPOL

A paper published @PNASNews today: "three current AI systems achieve a pass rate of at least 50% in a standard Turing test" The systems were GPT-4o, LLaMa-3.1, and GPT-4.5 All over 1-2 years old. https://www.pnas.org/doi/full/10.1073/pnas.2524472123

8:27 PM · May 19, 2026 · 10.4K Views
QUOTE POSTTuhin Chakrabarty#1050Tuhin Chakrabarty@TUHINCHAKR

🔥🔥🔥

Cameron JonesCameron Jones@camrobjones

Really excited that this is out in @PNASNews! We find that 2 LLMs (GPT-4.5 and LLama-3.1-405B) pass a 5 minute Turing test. As an update to our preprint we also find that GPT-5 and LLaMa pass a 15 minute test! 🧵

4:45 PM · May 19, 2026 · 1.9K Views
5:25 PM · May 19, 2026 · 616 Views