Gemma 4 is finally stable on llama.cpp On April 2nd, Google released Gemma 4. It had llama.cpp support on day one, but there were lots of bugs. Now, all issues have been fixed. E2B, E4B, 26B MoE, 31B Dense 31B ranks #3 on Arena AI, 26B ranks #6 Top-tier open-source models Use `--chat-template-file` to load the interleaved template. Recommend enabling `--cache-ram 2048`. Context length depends on VRAM. Last year, the best local model was the Llama 3.1 70B quantized version, barely usable. Now, Gemma 4 31B Q5 runs smoothly on Mac Studio, close to GPT-4 level. AI applications without relying on APIs are starting to have commercial viability. Data stays on your machine, costs are zero, and latency is extremely low. For solo entrepreneurs, local models are the real infrastructure. While competitors are paying API fees, your marginal cost is just electricity. Gemma 4 + llama.cpp = the optimal solution for local inference. Ready for production.