Changelog
August 26, 2025
- 📷 VLM support
- 🧠 LFM2-VL models now available in model library
- 📱 Edge SDK now accepts image content type as message inputs for LFM2-VL models (v0.5.0 and up)
- 💻 Laptop support for LFM2-VL models out-of-the-box
- 🏹 Test LFM2-VL models live in Apollo (v1.1.4 and up)
- 🛠️ Finetuning tools accept LFM2-VL checkpoints for high customization capabilities
- 📦 Model Bundling Service accepts LFM2-VL finetuned checkpoints for export to Edge SDK
- 🛠️ Finetuning tools
- Liger kernel now supported out-of-the-box
August 18, 2025
- 💻 Laptop support
- Native support for laptop and desktop computers for LEAP models via llama.cpp
- Best-in-class performance on AMD Ryzen™ chips
- Examples using language bindings for NodeJS and Python
- See more details here
- 🛠️ Finetuning tools
- Fixed a bug that caused training to hang unexpectedly
- 🤖 Android Edge SDK
- Fixed a bug related to model downloading
August 11, 2025
- 🤖 Android Edge SDK
- Added function-calling support to aid in agentic tool-calling
- Update model downloader query result format
- 🍎 iOS Edge SDK
- Added function-calling support to aid in agentic tool-calling
August 4, 2025
- 🤖 Android Edge SDK
- Added constrained generation support to aid in structured output generation
- 🍎 iOS Edge SDK
- Added constrained generation support to aid in structured output generation
- 📦 Model Bundling Service
- Fixed a bug regarding file upload
July 28, 2025
- 🛠️ Finetuning tools
- Initial release of finetuning tools for LFM2 models on ≤ 1 GPU (Google Colab notebooks) and > 1 GPU (
leap-finetune
) - See more details here
- Initial release of finetuning tools for LFM2 models on ≤ 1 GPU (Google Colab notebooks) and > 1 GPU (
- 📦 Model Bundling Service
- Use the Finetuning customization tools described above along with
leap-bundle
to develop and deploy finetuned LEAP models on the edge - CLI-based service to create model bundles for usage within the Edge SDK - currently supports any model architecture within the LEAP model library
- See more details here
- Use the Finetuning customization tools described above along with
- 🤖 Android Edge SDK
- Features
- Added
ModelLoadingOptions
andGenerationOptions
for more finegrained control over generation and loading options - Exposed model ID via
ModelRunner.modelId
- Exposed generation statistics via
stats
field onMessageResponse.Complete
- Added Model Downloader module to simplify model fetching in prototypes and development; see details here
- Added
- 🐛 Bug fixes
- Added Proguard rules to preserve inference engine class names and prevent obfuscation issues.
- If the generation content hits the maximum context length,
finishReason
field ofMessageResponse.Complete
will beEXCEED_CONTEXT
. If the prompt exceeds the context length,LeapGenerationPromptExceedContextLengthException
will be thrown.
- Features
- 🍎 iOS Edge SDK
- Features
- Added
ModelLoadingOptions
andGenerationOptions
for more finegrained control over generation and loading options - Exposed model ID via
ModelRunner.modelId
- Exposed generation statistics via
stats
field onMessageResponse.Complete
- Added
LeapModelDownloader
module to simplify model fetching in prototypes and development; see details here
- Added
- 🐛 Bug fixes
- If the generation content hits the maximum context length,
finishReason
field ofMessageResponse.Complete
will beEXCEED_CONTEXT
. If the prompt exceeds the context length,LeapGenerationPromptExceedContextLengthException
will be thrown.
- If the generation content hits the maximum context length,
- Features
Last updated on