MI300x DSV4-flash real world-ish performance testing

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1vdsmnk/deepseek_v4_iq3xxs_on_m1_ultra_128gb_16_toks_in/"> <img alt="DeepSeek V4 @ IQ3XXS on M1 Ultra 128GB-

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

<!-- SC_OFF --><div class="md"><p>Right from the oven with the <a href="https://github.com/jundot/omlx/commit/58b3ca549ab7aba075ecd5f1481911e01d819702">latest commit</a>:</p>

<!-- 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>Reports of the demise of coders may have been exaggerated.</p> </div><!-- SC_ON --> &#32; submitted by &#32; <a

<!-- 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>

<!-- SC_OFF --><div class="md"><p>Running oQ4e+MTP on oMLX 0.7.0, with still more to optimize.</p> <p>Prefill is 1,878 toks.</p> <p>I saw some other benchmarks below what id

<!-- SC_OFF --><div class="md"><p>Actually I'm not someone with particularly deep technical knowledge but I want to build a product, and instead of paying Claude a lot of money,