lukasonedge

4 ms·

1. 1.▲
   by lukasonedge 2mo ago

   all runs are cryptographically hashed so users can't tamper. Thermals and peak memory is logged as well. Example: https://computearena.ai/benchmarks/de2ccdb2-c0a9-431b-8e0a-3...
2. 2.▲
   by lukasonedge 2mo ago

   i didn't recognise some of the models on the frontpage
3. 3.▲
   by lukasonedge 2mo ago

   good overview of hardware and models
4. 4.▲
   State of Local AI Report(basecompute.co)

   3 points by lukasonedge 2mo ago|1 comments
5. 5.▲
   Show HN: Zero () friction local AI for Mac(basecompute.co)

   2 points by lukasonedge 3mo ago|0 comments
6. 6.▲
   by lukasonedge 3mo ago

   hey guys, we've optimised our inference engine to run Meta Muse Glimmer 30B faster than any other engine (llama.cpp etc) on Mac. Best performance on M5 Pro.
7. 7.▲
   Fastest Inference Meta Muse Glimmer 30B on Apple(basecompute.co)

   2 points by lukasonedge 3mo ago|1 comments
8. 8.▲
   6.4x faster than llama.cpp, 3.9x faster than MLX(basecompute.co)

   4 points by lukasonedge 4mo ago|2 comments