Qwen3-1.7B-litertlm-npu is an open-weight model from Norihiro Murata, released under Apache License 2.0. Its published files total 2.1 GB.
An ahead-of-time compiled.litertlm of Qwen/Qwen3-1.7B for the Qualcomm Hexagon NPU (v79, SM8750), made for the Android app Chiko - Local AI Hub.
Model Card
By Norihiro Murata, published under apache-2.0, revision b9eb732ca0ce.
An ahead-of-time compiled.litertlm of Qwen/Qwen3-1.7B for the Qualcomm Hexagon NPU (v79, SM8750), made for the Android app Chiko - Local AI Hub. litert-torch 0.9.4 generative/exporthf/experimental/npuexport with ai-edge-litert 2.2.0 and the Qualcomm AI Runtime 2.47.0.260601 compiler plugin, calibrated on Japanese chat-formatted prompts. The prefill/decode model is entirely DISPATCHOP (runs on the NPU); the embedder and KV-cache updates run on the CPU, the same layout as Google's official NPU builds. It needs the LiteRT Qualcomm dispatch library and QNN runtime libraries on the device. The quantized model was run on a PC CPU with litert-torch's executor and scored with a 30-question Japanese…
Read Norihiro Murata's full model card
Qwen3-1.7B — LiteRT-LM NPU build for Snapdragon 8 Elite (SM8750)
[!WARNING] Experimental, not yet run on a real device. This file compiles and its quantized model answers Japanese on a PC, but it has not been loaded on a Snapdragon 8 Elite NPU yet. Use it for testing only.
An ahead-of-time compiled .litertlm of Qwen/Qwen3-1.7B
for the Qualcomm Hexagon NPU (v79, SM8750), made for the Android app
Chiko - Local AI Hub.
| File | Qwen3-1.7B_qualcomm_sm8750_ekv1792.litertlm |
| SoC | Qualcomm SM8750 (Snapdragon 8 Elite) only |
| Context | 1792 tokens (prefill 128) |
| Quantization | static range A16W8 (dynamic_wi8_afp32 weights) |
| Size | 2,075,431,265 bytes |
| SHA-256 | 67db07c2ad9eeb986fee407acb9e5b587439296e28dd471c2f60fc1df1e319b7 |
How it was made
litert-torch 0.9.4 generative/export_hf/experimental/npu_export with
ai-edge-litert 2.2.0 and the Qualcomm AI Runtime 2.47.0.260601 compiler plugin,
calibrated on Japanese chat-formatted prompts. The prefill/decode model is
entirely DISPATCH_OP (runs on the NPU); the embedder and KV-cache updates run
on the CPU, the same layout as Google's official NPU builds. It needs the LiteRT
Qualcomm dispatch library and QNN runtime libraries on the device.
Quality check (PC, not on device)
The quantized model was run on a PC CPU with litert-torch's executor and scored with a 30-question Japanese evaluation set (rules + an LLM judge):
| Build | Japanese score |
|---|---|
| Float (before NPU quantization) | 64 |
| This file's quantized model, context 1024 | 60 |
| This file's quantized model, context 1536 | 58 |
License
Apache License 2.0, the same as the base model. Qwen3 is developed by the Qwen team, Alibaba Cloud.
Identity and Version
- Repository
- Norixp/Qwen3-1.7B-litertlm-npu
- Publisher
- Norihiro Murata
- Task
- Not stated by the source
- Modality
- Other
- Library
- litert-lm
- Parameters
- Not stated by the source
- Languages
- npu
- Revision
- b9eb732ca0ce84558a1268ac1617c3dac374b726
- First published
- 2026-10-04
- Last updated
- 2026-10-04
Files and Weights
3 files, 2.1 GB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| README.md | Documentation | 2.0 KB | — |
| Qwen3-1.7B_qualcomm_sm8750_ekv1792.litertlm | Other | 2.1 GB | 67db07c2ad9e |
| .gitattributes | Repository | 1.6 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
Released by Norihiro Murata through its official repository on Hugging Face. Read the license.
Built From
- Derived from Qwen/Qwen3-1.7B
- Quantized from Qwen/Qwen3-1.7B
Questions About Qwen3-1.7B-litertlm-npu
Can I use Qwen3-1.7B-litertlm-npu commercially?
Yes. Qwen3-1.7B-litertlm-npu is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.