It’ll all come crashing down soon enough.
Nvidia unveils new GPU designed for long-context inference
Submitted 18 hours ago by misk@piefed.social to technology@lemmy.zip
https://techcrunch.com/2025/09/09/nvidia-unveils-new-gpu-designed-for-long-context-inference/
brucethemoose@lemmy.world 16 hours ago
Doubling down on flash attention (my interpretation of this) is quite risky, as there are more efficient attention mechanisms seeping into bigger and bigger models.
Deepseek’s MLA is a start. Jamba is already doing hybrid GQA/Mamba attention, and a Qwen3 update is rumored to be using something exotic as well.
In English, this seems like they’re selling the idea of the software architecture not changing much, when that doesn’t seem to be the case.
nymnympseudonym@piefed.social 16 hours ago
Any favorites? What do you think about state space models?
brucethemoose@lemmy.world 11 hours ago
Jamba (hybrid transformers/space state) is a killer model folks are sleeping on. It’s actually coherent at long context, fast, has good world knowledge, even/grounded, and is good at RAG Its like a straight up better Cohere model IMO, and a no brainer to try for many long context calls.
TBH I didn’t try Falcon H1 much when it seemed to break at long context for me. I think most folks (at least publicly) are sleeping on hybrid SSMs because support in llama.cpp is not great. For instance, context caching does not work.
…Not sure about many others, toy models aside. There really aren’t too many to try.