GPT-2
Shares rank 42 of 72 language models by recommenders
Recommended as an early language model that was impressive for 2019 and foreshadowed later AI. Critics say its output was barely coherent and its withheld release was overhyped.
Recommended as an early language model that was impressive for 2019 and foreshadowed later AI. Critics say its output was barely coherent and its withheld release was overhyped.
37 people gave an opinion
- recommend it
- 12
- criticize it
- 25
- approval share
- 32%
Recommended in 6 of 13 months
- November 2025: recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: not recommended
- April 2026: recommended
- May 2026: not recommended
- June 2026: recommended
- July 2026: not recommended
- August 2026: recommended
- September 2026: recommended
- October 2026: recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 12 people recommend itNeeds 10 or more, met.
- 0.4 recommenders for every criticNeeds 2 to 1 or better, not met.
- Recommended in 6 of 13 monthsNeeds 4 or more, met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Impressive for its time
“Just stringing together a few sentences while maintaining logical coherence was very impressive at the time.”
Comment from 6 Jun 2026 in “What was your "Oh Shit" moment with GenAI?” “It was still impressive for 2019, but yeah, it was nothing to worry about.”
Comment from 8 Apr 2026 in “OpenAI says its new model GPT-2 is too dangerous to release (2019)” Foreshadowed later model behavior
“that generation of models was around the point where we were getting pleasantly surprised by the behaviors of models”
Comment from 16 Nov 2025 in “Vintage Large Language Models” “Obviously GPT-2 wasn’t great at it, but the writing was on the wall quite literally.”
Comment from 6 Jun 2026 in “What was your "Oh Shit" moment with GenAI?” Interesting despite being useless
“smaller versions of the model that were interesting but pretty useless”
Comment from 10 Jun 2026 in “GPT-2: Too Dangerous To Release (2019)” “Practically useless, but interesting.”
Comment from 11 Oct 2026 in “Artificial Insanity”
What people warn about
Each mark is one comment
Output barely coherent, Markov-chain-like
“What I'm getting from it looks like it could just be arbitrary sentence and paragraph fragments from the Internet, pasted together Markov-chain like.”
Comment from 28 Sep 2026 in “MicroLLM Lab – Try 7 tiny LLM's in the browser” “GPT-2 wasn't capable of anything other than imitating a stroke victim.”
Comment from 9 Sep 2026 in “I resigned from Anthropic today” “GPT-2 could do those things about as convincingly as a Markov chain.”
Comment from 9 Sep 2026 in “I resigned from Anthropic today” - Comment from 8 Apr 2026 in “OpenAI says its new model GPT-2 is too dangerous to release (2019)”
- Comment from 15 May 2026 in “How Claude Code works in large codebases”
- Comment from 16 Apr 2026 in “Claude Opus 4.7”
Withheld release seen as overhyped
“OpenAI's qualms weren't with making the dangerous models usable, they were with making the model usable without paying them.”
Comment from 8 Apr 2026 in “OpenAI says its new model GPT-2 is too dangerous to release (2019)” “when it was the first time OpenAI restricted access to the models citing "humanity is not ready for it"”
Comment from 22 May 2026 in “Project Glasswing: An Initial Update” “We cared about misuse right up until it became apparent that was profit to be had”
Comment from 16 Apr 2026 in “Claude Opus 4.7” - Comment from 17 Jun 2026 in “The hacker sent by Anthropic to calm the government's nerves about AI safety”
Far worse than newer models
“vanilla transformers like GPT 2 are ridiculously inefficient in comparison”
Comment from 7 Oct 2026 in “Claude Haiku 5.5” “it sucks much much worse than GPT-2 did with dungeon text adventures”
Comment from 11 Jan 2026 in “Ask HN: Is Programming as a Profession Cooked?” Called useless and unimpressive
“GPT-2 was effectively useless because it was only trained from links posted on Reddit”
Comment from 9 Jun 2026 in “GPT-2: Too Dangerous To Release (2019)” “smaller versions of the model that were interesting but pretty useless”
Comment from 10 Jun 2026 in “GPT-2: Too Dangerous To Release (2019)” “It was bullshit in 2019 it's bullshit now too.”
Comment from 9 Jun 2026 in “GPT-2: Too Dangerous To Release (2019)”
Other language models people recommend
The four most recommended other language models, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| Claude Opus | 3,331 | 2,746 | 55% | 13 | 4 mo156 owners | — |
| OpenAI GPT | 1,205 | 987 | 55% | 13 | 6 mo47 owners | — |
| Claude Sonnet | 854 | 586 | 59% | 13 | 6 mo46 owners | — |
| DeepSeek V4 Flash | 828 | 223 | 79% | 8 | 1 mo24 owners | Approved |
| GPT-2This page | 12 | 25 | 32% | 6 | — | — |
What people said
37 of the 39 comments that mention it, newest first. Each links to the original comment.
“Practically useless, but interesting.”
“recall GPT-2 was too dangerous to release”
“vanilla transformers like GPT 2 are ridiculously inefficient in comparison”
“it's about the same level of being terrible”
“What I'm getting from it looks like it could just be arbitrary sentence and paragraph fragments from the Internet, pasted together Markov-chain like.”
“GPT-2 124M is terrible...”
“GPT2 was absolutely unusable as a classifier.”
“Fast, cheap, and open weights!”
“came back with an awesome syntax and set of phenome rules which made an awful lot of sense to me”
“OpenAI said GPT-2 was too powerful to release”
“GPT-2 could do those things about as convincingly as a Markov chain.”
“GPT-2 wasn't capable of anything other than imitating a stroke victim.”
“Able to generate realistic spam at arbitrary volume.”
“Amodei also fear-mongered GPT-2”
“GPT 2.5 was too dangerous, for example.”
“Look how much worse society got that directly traces back to releasing GPT-2.”
“GPT-2 was a gibberish generator.”
“OpenAI is also guilty of excessive fear mongering”
“smaller versions of the model that were interesting but pretty useless”
“It was bullshit in 2019 it's bullshit now too.”
“GPT-2 was effectively useless because it was only trained from links posted on Reddit”
“Just stringing together a few sentences while maintaining logical coherence was very impressive at the time.”
“Obviously GPT-2 wasn’t great at it, but the writing was on the wall quite literally.”
“create jokes that were original and kinda funny”
“when it was the first time OpenAI restricted access to the models citing "humanity is not ready for it"”
“That model could barely string a sentence together.”
“that output is barely better than a Markov chain”
“We cared about misuse right up until it became apparent that was profit to be had”
“1.5B and insanely under-trained”
“GPT-2 wasn't that great of a model”
“OpenAI's qualms weren't with making the dangerous models usable, they were with making the model usable without paying them.”
“the model went off the rails with some rhetoric entangling "Trump", "Russia", "China", "CIA", "Voting"”
“It was still impressive for 2019, but yeah, it was nothing to worry about.”
“the thing could barely make full grammatical sentences”
“GPT-2, o1, Opus...been here so many times.”
“it sucks much much worse than GPT-2 did with dungeon text adventures”
“that generation of models was around the point where we were getting pleasantly surprised by the behaviors of models”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.