GPT-4
No written verdict yet. Here is what commenters said.
By version sentiment of people who named that release, latest release first
Versions with at least 5 people. 0 people didn't name a version. Small samples swing a lot, so read the counts too. Compare all models →
Why people recommend it
“GPT4 and GPT5 were more than enough supplementation”
Show 3 more
“This study suggests that GPT-4 models could be leveraged as a support tool in mental health telephone triage”
“GPT4, when it could do a translation that would take a considerable human effort”
“the best one so far is GPT4.1-Mini”
Caveats and dissent
“Prompt rules alone had GPT-4o inventing IDs to satisfy them.”
“ChatGPT 4o got a lot of people addicted.”
Show 2 more
“gpt4.0 or whatever would often make hideous abominations of legs and mouths”
“GPT 4o is disgusting, just horrible saccharine nonsense”
Often compared with
3 moved from GPT-4 to GPT-5
1 moved from GPT-4 to Claude
Who recommends it in their own words
Threads where it came up
- OpenAI's GPT-4 Discontinuation: Consumer Fraud and Regulatory Scrutiny 6 mentions
- Good Riddance, 4o 5 mentions
- Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT 3 mentions
- Karpathy's Pelican 2 mentions
- "No, I swear I wrote this." 2 mentions
- Reallocating $100/Month Claude Code Spend to Zed and OpenRouter 2 mentions
Also written as GPT-4, GPT-4.5, GPT-4.1, GPT-4o-mini. Data: gpt-4.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 48
“GPT-3 vs GPT-4 was practically a 10-100x jump”
“I remember the same doom cycle during my PhD when GPT 4 came out”
“Prompt rules alone had GPT-4o inventing IDs to satisfy them.”
“We came a long way from 4o-mini”
“GPT4 and GPT5 were more than enough supplementation”
“This study suggests that GPT-4 models could be leveraged as a support tool in mental health telephone triage”
“ChatGPT 4o got a lot of people addicted.”
“GPT-4 (!) has, in studies comparing multiple models with humans for creativity, beaten the mean human.”
“gpt4.0 or whatever would often make hideous abominations of legs and mouths”
“I doubt anyone could give a not coding example where it’s meaningfully better with current frontier than 4o.”
“performance competitive with GPT-4o”
“GPT 4o is disgusting, just horrible saccharine nonsense”
“GPT 4o likes writing poetry with bees in it, for some reason.”
“This wasn't a new model but updates to models backed by numbers being better can make the model worse”
“the gains compared to the model size didn't seem that great”
“GPT4, when it could do a translation that would take a considerable human effort”
“It roughly compares with GPT-4.1”
“The current estimate is about $0.40 per million tokens for gpt4 equivalent model.”
“the best one so far is GPT4.1-Mini”
“GPT-4 (original API):”
As a pt-BR speaker from across the pond: https://soberania.ai/ Similar waste.
“You can run GPT-4 level models locally today for almost nothing in compute cost.”
“I've seen figures that suggest GPT-4 was 1.8T parameters”
“GPT-5 mini (medium) used 3x the energy of GPT-4o (Aug)”
“GPT-4o extracts vocabulary from the transcript and generates cards”
“GPT-4o, GPT-4.1, and GPT-5-mini”
“GPT-4o, GPT-4.1, and GPT-5-mini”
“the same was roughly true with chatgpt 4o”
“Multimodal LLMs like gpt-4o-transcribe are way better than traditional ASR”
“Mistral Large 3 and GPT-4.1 will initially agree with the narrator, while Gemini will disagree.”
“they all fail to consistently follow their own guardrails”
“shows that RLHF destroyed its calibration”
“it scores them 1-10 with GPT-4o-mini”
Yeah we all converge to the same workflow, in my ai coding agent I'm working on now, I've added an "index" tool that uses tree-sitter to compress and show the AI a skeleton of a code file. Here's the implementation for the interested: https://github.com/tontinton/maki/blob/main/maki-code-index%2Fsrc%2Flib.rs
“I use Flash for code maps and auto-context, and GPT-4.5 or Opus for coding”
“It uses gpt-4o-mini with enforced JSON mode”
“It worked well enough that the novels were better than the median novel aimed at my son's age group”
The main differences are transcription quality and what happens after the transcript is generated. Utter uses GPT-4o Transcribe by default for cloud transcription, and in my experience it’s best in class. The gap is most obvious on names, niche terminology, and technical vocabulary. I use it a lot for prompting coding agents, and I've found Apple’s built-in dictation and most other apps don't come close in terms of a
I have nothing to contribute but speculation based on my intuition, but IMO RLHF (or rather human preference modeling in general, including the post-training dataset formatting) is a relatively small factor in this, RL-induced mode collapse is much bigger one. Take a look at the original DeepSeek R1 Zero, the point of which was to train a model with very little human preference, because they've been on a budget and h
“GPT-4o's second-most-overused word”
“GPT-4.1 generates commentary”
“GPT-4.1 crushes certain coding tasks”
“4.5 was also good at this”
“I use it via my ChatGPT-Pro subscription”
“I used https://openrouter.ai/openai/gpt-4.1 for grammar checking, it was great.”
“I use Azure GPT 4.1 for limited llm tasks”
“4.5 was easily the best conversationalist I've seen”
“Last time I used GPT-4.5 to analyze blood results it gave different output”