Jon

[Jon @JonJonloretz]

@CoinMarketCap Hacked means TP'ed.

[Jon @JonJonloretz]

@CoinMarketCap That is the business model. Every exchange gets hacked every insider gets a clean exit. There is no such thing as a crypto hack.

[Jon @JonJonloretz]

@benjamincowen 62k incoming

[Jon @JonJonloretz]

@Alibaba_Qwen @NVIDIARTXSpark Every time Qwen drops a model, the open-source community wins. Qwen3.8-35B-A3B is the next chapter we're all waiting for. We're grateful, we're ready and we appreciate everything you do.

[Jon @JonJonloretz]

@AntLingAGI Nobody is reacting. Nobody is interested. Join the party in the 30b a3b and you'll see.

[Jon @JonJonloretz]

@OfficialLoganK Cancel gemini pro. Gemma = Android

[Jon @JonJonloretz]

@finkd Nokia called Android "peeing in your pants" in 2010. Now Symbian is gone and Android runs the world. Open source AI is the Android of LLMs and the future is wide open.

[Jon @JonJonloretz]

@finkd In 2010, Nokia's VP called Android "peeing in your pants to stay warm." A few years later, Symbian was dead and Android ruled the world. Open source AI is the Android of LLMs and the same pattern is playing out. Great move by Meta!

[Jon @JonJonloretz]

@alexandr_wang MOE FTW!

[Jon @JonJonloretz]

@OfficialLoganK Gemma

[Jon @JonJonloretz]

@deepgrove_ai when gguf for llama.cpp?

[Jon @JonJonloretz]

@QwenDevs 30b/35b a3b please?

[Jon @JonJonloretz]

@Eschalabs gguf, llama.cpp

[Jon @JonJonloretz]

@osanseviero Just beat qwen3.6 35b. End of the discussion.

[Mark Zuckerberg @finkd]

Open source is a positive and important force for both empowering people and preventing centralization. Proud to support this.

[Jon @JonJonloretz]

@Kimi_Moonshot Nice. But we want that 30B A3B open source energy. Drop it and watch the community explode!

[Jon @JonJonloretz]

@ChujieZheng Thanks, Alibaba! Big news.

But can we talk about the elephant in the room?

The models have developed this extreme overthinking habit:

"Wait, I will answer now, wait, but wait... there's more! But wait...."

Can we fix this? We need models to just answer instead of looping.

[Jon @JonJonloretz]

@above_spec Can you also show the benchmarks using the single RTX 5060 ti?

[Jon @JonJonloretz]

@Baidu_Inc MOE A3B please?

[Jon @JonJonloretz]

@ZyphraAI @AMD 30(ish)B A3B parameters is the sweet spot for open source LLMs.