To my understanding, Reasoning Model refers to models that have been explicitly fine-tuned to perform chain-of-thought reasoning effectively. Something analogous to DeepSeek’s GRPO. For instance, even if I told GPT-3 to show its work, it would not be a reasoning model.
As for LLM, it’s admittedly on the path to being a holdover term, but not for lack of effort. Some people tried to make VLM (Vision-Language Model) more common, but I’ve mainly heard it used in the context of lightweight models used to graft vision capability onto systems driven by single-modality LLMs.
To my understanding, Reasoning Model refers to models that have been explicitly fine-tuned to perform chain-of-thought reasoning effectively. Something analogous to DeepSeek’s GRPO. For instance, even if I told GPT-3 to show its work, it would not be a reasoning model.
As for LLM, it’s admittedly on the path to being a holdover term, but not for lack of effort. Some people tried to make VLM (Vision-Language Model) more common, but I’ve mainly heard it used in the context of lightweight models used to graft vision capability onto systems driven by single-modality LLMs.