No 'GPT-5 level' model, a huge across-the-board leap by community consensus, may appear throughout 2025.

Judge at the end of 2025, against the claim's own definition. A 'GPT-5 level' model means a huge, across-the-board quantum leap forward over prior frontier models, as judged by community consensus. This is comparable to the jump from GPT-3 to GPT-4, not an incremental improvement or progress on a few benchmarks only. A model's name, including one called 'GPT-5', does not settle the question; only the consensus judgement of the leap does. The claim is strongly hedged ('could'), so resolve it on whether the possibility it raises came about. TRUE: no model released in 2025 is regarded by broad community consensus (researchers, independent evaluators, major commentators) as a huge across-the-board leap. MOSTLY TRUE: no consensus leap occurred, but at least one 2025 model is argued by a notable minority to have been one. MIXED: the community is roughly evenly divided on whether a 2025 model was such a leap. MOSTLY FALSE: a majority, though not a clear consensus, views a 2025 model as such a leap. FALSE: there is clear community consensus that a model released in 2025 was a huge across-the-board quantum leap.

Set before the research began.

1. E1 GPT-5 Benchmarks and Analysis Artificial Analysis · 7 August 2025

   “While High sets a new standard, the increase over o3 is not comparable to the jump from GPT-3 to GPT-4 or GPT-4o to o1.”

2. E2 GPT-5's modest gains suggest AI progress is slowing down New Scientist · 13 August 2025

   “It has improved on GPT-4, but the difference for many benchmarks is smaller than the leap from GPT-3 to GPT-4.”

3. E3 GPT-5 failed the hype test The Verge · 15 August 2025

   “The general consensus is that although GPT-5 wasn’t as significant of an advancement as people expected, it offered upgrades in cost and speed, plus fewer hallucinations”

4. E4 Gemini 3 just got *scary* good readray.ai (LilysAI video note) · 26 November 2025

   “This analysis confirms that Gemini 3 is a massive, non-incremental leap forward, crushing competitors across critical benchmarks like Vending Bench 2 and Humanity's Last Exam.”

The evidence shows independent sources did not regard GPT-5 as a GPT-3-to-GPT-4-scale leap (E1, E2, E3), but the only item on Gemini 3 is a single auto-generated video note arguing it was such a leap (E4). The evidence does not show how the wider community judged Gemini 3, so it cannot settle whether any 2025 model was seen by consensus, or by a notable minority, as a huge across-the-board leap.