/Tech3h ago

LightCell Energy's Danielle Fong argues GPT-4.5 was restricted by high inference costs, while j⧉nus blames failed post-training

Fong contested claims that the model had flawed pre-training

774015.4K

#674

Original post

Danielle Fong 🔆@DanielleFong#756inTech

gpt4.5 was amazing. it wasn't a botched pretrain. it was just explictly explicitly expensive to serve

JMB 🧙‍♂️@jmbollenbacher

OpenAI wont catch Anthropic because they forgot how to pretrain.

GPT-4.5 was supposed to be their best model ever. It flopped. Botched pretrain. GPT-5.5 was supposed to be GPT-6, but it was incremental over 5.4. Failure of ambition or execution.

OpenAI can't make a Mythos/Fable

11:53 AM · Jun 19, 2026 · 4.7K Views

Sentiment

Some users criticized GPT-4.5's post-training as botched after its founder defended the model as effective but expensive to serve.

Pos

0.0%

Neg

100.0%

1 comments with sentiment.

Cluster Engagement

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

Posts from X

Most Activity

VIEWS1.2KBOOKMARKS1LIKES31REPLIES3

j⧉nus@repligate

@DanielleFong it was a botched posttrain

Danielle Fong 🔆@DanielleFong

gpt4.5 was amazing. it wasn't a botched pretrain. it was just explictly explicitly expensive to serve

3h1.2K311