RecruitLingo

Terms that look alike but mean genuinely different things.

Chrome extensionsoon

BERT vs. GPT-style LLMs

New

Both are transformer models, so 'transformer experience' can mean either. BERT-style models read text and turn it into labels or numbers: classification, search, named entities. They are small enough to fine-tune on a company's own data and run cheaply. GPT-style models, the large language models behind ChatGPT and Claude, generate text. Much of the NLP work before 2023 was BERT-style and most since is built on LLMs, so the two point to different eras and skills.

How to tell them apart on a resume

BERT-style (reads and labels)

BERT, RoBERTa, DistilBERT, sentence-transformers, text classification, NER, fine-tuning on labelled data, Hugging Face Transformers.

GPT-style LLMs (generate)

GPT, Claude, Gemini, Llama, prompting, RAG, LLM APIs, agents, text generation, fine-tuning with LoRA.

The question that settles it

“Were you training models that classify or label text, like BERT, or building with models that generate text, like GPT or Claude?”

Each term in plain English

  • BERT

    A Google language model from 2018 that reads a whole sentence at once to understand each word in context. Unlike chat models it doesn't write text; it's used for classifying text, picking out names and entities, and search.

    Ask “What did you fine-tune BERT for, how much labelled data did you have, and why not use an LLM instead?”

  • LLM

    A large AI model (like GPT or Claude) trained on huge amounts of text, able to understand and generate human-like language.

    Ask “What have you actually built with an LLM — prompting, fine-tuning, evaluation, or the surrounding application?”

  • Frontier Models (GPT, Claude, Gemini)

    The most capable AI models, built by a few well-funded labs and used through their apps or paid APIs: OpenAI's GPT models (behind ChatGPT), Anthropic's Claude and Google's Gemini.

    Ask “Which models did you build with through the API, and how did you choose between them for your use case?”

  • Transformer

    The neural network architecture (introduced in 2017) behind nearly all modern LLMs, notable for processing entire sequences of text at once via a mechanism called 'attention.'

    Ask “Can you explain, at a high level, what the 'attention' mechanism in a transformer is actually doing?”

Open the full tool for the other look-alike pairs, role profiles, and the JD decoder.