nerfed
In plain wordsGaming slang for something made weaker by an update; applied to AI models that users believe got worse after a provider changed them, often without saying so.
Of an AI model or product: weakened by its provider, especially through an unannounced update — shorter or lazier answers, tighter usage limits, a smaller or more heavily quantizedquantization, n.: Storing an AI model's numbers with fewer digits, like rounding prices to the nearest dollar, so the model takes less memory and runs faster, usually with only a small loss of accuracy. model behind the same name, or reduced reasoning effort. Suspected 'stealth nerfs' are a recurring theme in AI communities because hosted models can change without notice and output quality varies from session to session. Providers sometimes acknowledge regressions — OpenAI did so when users complained in late 2023 that GPT-4 had become 'lazy' — but many claims reflect randomness, harder tasks or rising expectations. The difficulty of settling such disputes has encouraged independent tracking of model performance over time, pinned version identifiers and public changelogs.
Same prompts, same repo, half the quality — I think they nerfed it to save compute.
Origin
From gaming slang of the 1990s, after the soft-foam Nerf toys: a weapon or character 'nerfed' by a patch was made less effective, a usage widely traced to the online game Ultima Online (1997). Applied to AI models from 2023, as ChatGPT users compared answers before and after updates.
Usage note
Not to be confused with NeRF, the neural radiance fieldneural radiance field, n.: A way of capturing a real scene from a set of photos as a small neural network that can then show the scene from any new viewpoint, as if moving a camera through it. method for 3D scene reconstruction.
Related
- lobotomizedlobotomized, adj.: Slang complaint that an AI model has been made duller, more cautious or less capable — usually by safety restrictions or cost-cutting — as if it had undergone brain surgery.
- quantizationquantization, n.: Storing an AI model's numbers with fewer digits, like rounding prices to the nearest dollar, so the model takes less memory and runs faster, usually with only a small loss of accuracy.
- vibe checkvibe check, n.: Trying out a new AI model informally — chatting with it, asking your favorite test questions — to get a gut feeling for how good it is, rather than measuring it with formal tests.
- it's over, we're so backit's over, we're so back, phr.: A meme describing the mood swings of the AI world: one day bad news means 'it's over', the next a new model or result means 'we're so back'.
- benchmaxxingbenchmaxxing, n.: Slang for tuning an AI model to score as high as possible on well-known tests, even when this does not make it genuinely better at real-world tasks.
Mentioned in
lobotomizedlobotomized, adj.: Slang complaint that an AI model has been made duller, more cautious or less capable — usually by safety restrictions or cost-cutting — as if it had undergone brain surgery.