ran 150+ benchmarks across a bunch of macs, here's what we found

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1tq7b75/i_ran_some_benchmarks_using_omlx_tool_i_know_not/"> <img alt="I ran some benchmarks using oMLX tool

<!-- SC_OFF --><div class="md"><p>TLDR Although technically Qwen 3.5 397B Q8_0 fits on my server, and can process a one-off prompt, so far I’ve not found it to be practical for

<!-- SC_OFF --><div class="md"><p>Source: <a

<!-- SC_OFF --><div class="md"><p>So i recently updated LMstudio after a long pause and updated my llama.cpp runtimes too.. i was shocked.. i thought maybe something like

<!-- SC_OFF --><div class="md"><p>I’ve managed to get it to run and, with some patches to llama.cpp, I get ~8 tok/s decode and ~30 tok/s prefill on a M1 Max 64GB MacBook.</p>

<table> <tr><td> <a href="https://www.reddit.com/r/singularity/comments/1v2l6sm/gemini_36_flash_benchmarks/"> <img alt="Gemini 3.6 Flash benchmarks"

<table> <tr><td> <a href="https://www.reddit.com/r/ChatGPT/comments/1x5lxc9/pelican_benchmark_with_gpt_gemini_claude/"> <img alt="Pelican benchmark with GPT / Gemini / Claude"

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1siv4qy/amorphousmemorymark_which_benchmarks_memory/"> <img alt="AmorphousMemoryMark (which benchmarks