Is the model actually a Merge?
Curious about the wording in the model card "Mistral Medium 3.5 is our first flagship merged model." -- This is a very strange statement, the blog extrapolates: "Mistral Medium 3.5, a new flagship model that merges instruction-following, reasoning, and coding into a single 128B dense model"
Could you clarify if this is just strangely worded way of saying "Model that does instruct, reasoning, and coding" or is this saying the model is literally a merge of other models (And if so, which models?)
Hey, it is a model that does instruct, reasoning, and coding as a standalone.
It is not literally a merge of multiple other models.
Due to its capabilities, it replaces the previous models that used to have separate concerns: Mistral Medium 3.x for instruct, Devstral 2 for Coding and Magistral 1.x for reasoning. This is what we meant by "merging", instead of having 3 different models, one has all capabilities.
Thank you for the clarification 👍