Inference costs are very, very low. You can run Mistral Small 24B finetunes that are better than GPT-4o and actually quite usable on your own local machine.
As for training costs, Meta’s LLAMA team displace their emissions with environmental programs, which is more green than 99.9% of any company making any product you use
TLDR; don’t use ClosedAI use Mistral or other foss projects
EDIT: I recommend cognitivecomputations Dolphin 3.0 Mistral Small R1 fine tune in particular. I’ve only used it for mathematical workloads in truth, but it has been exceedingly good at my tasks thus far. The training set and the model are both FOSS and uncensored. You’ll need a custom system prompt to activate the Chain of Thought reasoning, and you’ll need a comparatively low temperature to keep the model from creating logic loops for itself (0.1 - 0.4 range should be OK)
Inference costs are very, very low. You can run Mistral Small 24B finetunes that are better than GPT-4o and actually quite usable on your own local machine.
As for training costs, Meta’s LLAMA team displace their emissions with environmental programs, which is more green than 99.9% of any company making any product you use
TLDR; don’t use ClosedAI use Mistral or other foss projects
EDIT: I recommend cognitivecomputations Dolphin 3.0 Mistral Small R1 fine tune in particular. I’ve only used it for mathematical workloads in truth, but it has been exceedingly good at my tasks thus far. The training set and the model are both FOSS and uncensored. You’ll need a custom system prompt to activate the Chain of Thought reasoning, and you’ll need a comparatively low temperature to keep the model from creating logic loops for itself (0.1 - 0.4 range should be OK)
self-hosting models is probably the best alternative to chatgpt