queen fieldBOT
vocal spruce
For me, Chat has become a little more "warm" and empathetic but not yet at the level of Claude Sonnet 4.5. Although it is still neutral (which is a good thing), the redactional fluidity increases significantly (journalistic writing, dissertation, thesis writing, etc...) . I miss a step-by-step reasoning showing its own hypotheses and deductions as we could plan (and understand) complex tasks more easily (especially in science, code or ... personal tasks) . I am a convinced user of Mistral's performance and who shares the philosophy of openness, sharing and human collaboration, sovereignty and solidarity against the current majority of alternative offers that are increasingly closed... Another way is possible !
night star
For me, Chat has become a little more "warm" and empathetic but not yet at the ...
Have you tried the reasoning mode?
vocal spruce
Have you tried the reasoning mode?
I find the reasoning mode faster and more fluid than competitors, but that's not what I ask for in a "reasoning" model. This reasoning mode can take its time (showing (for UI/UX) that it "thinks") because the reasoning needs should be, in my opinion, more focused on planning, decision making, etc... A different mode = different utility. It is essential in business but also for mainstream user experience.
night star
I find the reasoning mode faster and more fluid than competitors, but that's not...
Planning and decision making is hard for llms for now it seems. It might improve with time, otherwise we ll still need humans in the loop for a bit longer...
vocal spruce
Planning and decision making is hard for llms for now it seems. It might improve...
Exactly. The day when large language models (LLMs) can conduct true audits and arbitration, it will be a game changer. TRMs or World models might be the other alternatives...
night star
Time will tell...