Thank you !!
#13
by pawarshardul - opened
@reflection team thank you for open weights !! And i think you are the only major llm lab which shows how much quantization affects the model -- i never seen any of the llm provider showing the benchmarks across bf16,fp8 and nvfp4 !! Really appreciable!!
Agreed, thank you! Breath of fresh air from a United States team!
Using it now I ain't going to lie this is impressive what are the chances we will see smaller models from you like a flash model say 120B MoE and an edge model 20 to 35B
Hey there - definitely on the roadmap. tysm for all of the thank yous. It's a a labor of love for this team so we are really grateful you are here and using our models. Cheers!