shaz0x

5 ms·

1. 1.▲
   by shaz0x 5mo ago

   On mobile the Q4 vs Q6 tradeoff flips. Gemma 4 E2B at Q4_K_M barely fits in RAM on a 6GB Android, so Q6 isn't on the table. In practice the Q4 hit shows up in tool-call reliability more than general reasoning, which is usually fine for
2. 2.▲
   by shaz0x 5mo ago

   Went through the SDK docs before asking. On RN/Expo specifically, does Fabric run inside a Bare worklet with IPC back to Hermes, or drop into a native module the way llama.rn does via JNI and llama.cpp? Perf and memory footprint would
3. 3.▲
   by shaz0x 5mo ago

   Even Gemma 4 E2B is more useful than you'd think if you give it the right harness. I've been running it on Android via llama.rn and it handles function calling natively — the model outputs structured tool calls without any prompt