Sometimes I miss the cheerish and sycophant GPT-4o. There, I said it.
LLMS
-
GPT-5 Admits Uncertainty Instead of Hallucinating
By
–
GPT-5 admits it "doesn't know" an answer! This is one of the huge improvements over previous models: instead of hallucinating, it lets you know its limits.
-
Grok 4 Heavy Tops AI Rankings But Remains Too Expensive
By
–
grok 4 heavy est la meilleure IA dispo aujourd'hui mais c'est tellement cher que peu de monde l'utilise
-
GPT-5 User Preference Majority Over ChatGPT Analysis
By
–
I suspect this is right. And I wouldn’t be surprised if the vast majority of the 700M users of ChatGPT already greatly prefer GPT-5 & that the opinion on X is not reflective of the typical experience. (Which doesn’t mean that the issues identified here aren’t very real)
-
Google and DeepMind Impress with Recent AI Releases
By
–
Google est celui qui m'a le plus impressionné jusqu’à présent cette année. Demis Hassabis : fait le bilan de ce qu'ils ont sorti les 2 dernières semaine. C'est juste dingue. En 2014, Google a racheté DeepMind pour 650 millions de dollars. (Un prix qui a déjà été rentabilisé
-
Custom Qwen Inference Servers Hardware Cost Analysis
By
–
Well we have our own inference servers and custom config for qwen, so it's really the cost of that hardware I guess.
-
Grok-4 vs Grok-3: Quality Comparison Analysis
By
–
There is indeed a big difference between grok-3 and grok-4. But… it isn't because grok-4 is really good.
-
Deep Research Model Outdated Compared to GPT 5
By
–
Looks like Deep Research is still using an older model? GPT 5 knows about FastHTML (which is a year old), but if I select "Deep research" then it doesn't.
