Model reference · open weights
react-native-executorch-lfm-2.5 is an open-weight language model from software-mansion. Built for phones (ExecuTorch). It does not run on a GPU server.
What it is
| Released by | software-mansion |
|---|---|
| Released | 2026-04-01 |
From the model card
This repository hosts the lfm-2.5 models exported for the
React Native ExecuTorch
library as ExecuTorch .pte programs, ready to run on device.
| Path | Backend | Precision |
|---|---|---|
1_2b/mlx/lfm_2_5_1_2b_mlx_int4.pte | mlx | 4w |
1_2b/xnnpack/lfm_2_5_1_2b_xnnpack_8da4w.pte | xnnpack | 8da4w |
1_2b/xnnpack/lfm_2_5_1_2b_xnnpack_fp16.pte | xnnpack | fp16 |
350m/mlx/lfm_2_5_350m_mlx_int4.pte | mlx | 4w |
350m/xnnpack/lfm_2_5_350m_xnnpack_8da4w.pte | xnnpack | 8da4w |
350m/xnnpack/lfm_2_5_350m_xnnpack_fp16.pte | xnnpack | fp16 |
vl_1_6b/mlx/lfm_2_5_vl_1_6b_mlx_int4.pte | mlx | int4 |
vl_1_6b/mlx/lfm_2_5_vl_1_6b_mlx_int8.pte | mlx | int8 |
vl_1_6b/vulkan/lfm_2_5_vl_1_6b_vulkan_8da4w.pte | vulkan | 8da4w |
vl_1_6b/xnnpack/lfm_2_5_vl_1_6b_xnnpack_8da4w.pte | xnnpack | 8da4w |
vl_450m/mlx/lfm_2_5_vl_450m_mlx_int4.pte | mlx | int4 |
vl_450m/vulkan/lfm_2_5_vl_450m_vulkan_8da4w.pte | vulkan | 8da4w |
vl_450m/xnnpack/lfm_2_5_vl_450m_xnnpack_8da4w.pte | xnnpack | 8da4w |
1_2b/mlx/config.json 1.2 kB
1_2b/mlx/lfm_2_5_1_2b_mlx_int4.pte 849 MB
1_2b/tokenizer.json 4.5 MB
1_2b/tokenizer_config.json 92.4 kB
1_2b/xnnpack/config.json 2.1 kB
1_2b/xnnpack/lfm_2_5_1_2b_xnnpack_8da4w.pte 759 MB
1_2b/xnnpack/lfm_2_5_1_2b_xnnpack_fp16.pte 2.4 GB
350m/mlx/config.json 1.2 kB
350m/mlx/lfm_2_5_350m_mlx_int4.pte 308 MB
350m/tokenizer.json 4.5 MB
350m/tokenizer_config.json 2.9 kB
350m/xnnpack/config.json 2.1 kB
350m/xnnpack/lfm_2_5_350m_xnnpack_8da4w.pte 265 MB
350m/xnnpack/lfm_2_5_350m_xnnpack_fp16.pte 806 MB
config.json 29 B
vl_1_6b/mlx/config.json 1.7 kB
vl_1_6b/mlx/lfm_2_5_vl_1_6b_mlx_int4.pte 1.4 GB
vl_1_6b/mlx/lfm_2_5_vl_1_6b_mlx_int8.pte 2.0 GB
vl_1_6b/tokenizer.json 4.5 MB
vl_1_6b/tokenizer_config.json 3.2 kB
vl_1_6b/vulkan/config.json 2.1 kB
vl_1_6b/vulkan/lfm_2_5_vl_1_6b_vulkan_8da4w.pte 2.4 GB
vl_1_6b/xnnpack/config.json 2.1 kB
vl_1_6b/xnnpack/lfm_2_5_vl_1_6b_xnnpack_8da4w.pte 2.3 GB
vl_450m/mlx/config.json 1.0 kB
vl_450m/mlx/lfm_2_5_vl_450m_mlx_int4.pte 376 MB
vl_450m/tokenizer.json 4.5 MB
vl_450m/tokenizer_config.json 3.2 kB
vl_450m/vulkan/config.json 2.1 kB
vl_450m/vulkan/lfm_2_5_vl_450m_vulkan_8da4w.pte 626 MB
vl_450m/xnnpack/config.json 2.1 kB
vl_450m/xnnpack/lfm_2_5_vl_450m_xnnpack_8da4w.pte 619 MB
These files are published for the ExecuTorch v1.4.1 runtime. ExecuTorch gives no forward compatibility guarantee, so an older runtime may fail to load them.
To use them in React Native ExecuTorch, pass the model constant shipped in the library's model registry to the corresponding task pipeline. See the documentation.
To load these files in your own ExecuTorch runtime, read the compatibility note first.
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.
How it works
Running it
Built for phones (ExecuTorch). It does not run on a GPU server.