Qwen3 4B Instruct 2507 for RK3588 (RKLLM 1.2.1) This repository contains versions of Qwen/Qwen3 4B Instruct 2507 converted to run on the RK3588 NPU. Compatible with RKLLM version: 1.2.1 Available Models Quantization Optimization Hybrid Ratio Context Filename w8a8 g128 0 0.0 4k Qwen3 4B Instruct 2507 rk3588 w8a8 g128 opt 0 hybrid ratio 0.0 4k.rkllm w8a8 g128 1 0.0 4k Qwen3 4B Instruct 2507 rk3588 w8a8 g128 opt 1 hybrid ratio 0.0 4k.rkllm w8a8 g128 0 0.2 4k Qwen3 4B Instruct 2507 rk3588 w8a8 g128 opt 0 hybrid ratio 0.2 4k.rkllm w8a8 g128 1 0.2 4k Qwen3 4B Instruct 2507 rk3588 w8a8 g128 opt 1 hybrid ratio 0.2 4k.rkllm w8a8 g128 0 0.4 4k Qwen3 4B Instruct 2507 rk3588 w8a8 g128 opt 0 hybrid ratio 0.4 4k.rkllm w8a8 g128 1 0.4 4k Qwen3 4B Instruct 2507 rk3588 w8a8 g128 opt 1 hybrid ratio 0.4 4k.rkllm w8a8 0 0.0 4k Qwen3 4B Instruct 2507 rk3588 w8a8 opt 0 hybrid ratio 0.0 4k.rkllm w8a8 1 0.0 4k Qwen3 4B Instruct 2507 rk3588 w8a8 opt 1 hybrid ratio 0.0 4k.rkllm w8a8 0 0.2 4k Qwen3 4B Instruct 2507 rk3588 w8a8 opt 0 hybrid ratio 0.2 4k.rkllm w8a8 1 0.2 4k Qwen3 4B Instruct 2507 rk3588 w8a8 opt 1 hybrid ratio 0.2 4k.rkllm w8a8 0 0.4 4k Qwen3 4B Instruct 2507 rk3588 w8a8 opt 0 hybrid ratio 0.4…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy