Release v0.8.7

• Fixed critical crashes
• Fixed model export bug
• Improved thinking parser for reasoning models
• Added sliders for model parameters settings
• Removed default system prompt
• Set default mirostat to 0
• Set default context size to 4096

Release v0.8.6

• Improved RAG for local models
• Added native file upload support for remote models
• Added new tab switcher layout for tablets
• Improved chat history title generation
• Added ability to pin chats

v0.8.5

• Improved server logs
• Bug fixes

Available on App Store:

• Migrated to self-hosted @inferrlm/react-native-mlx, added support for Qwen 3.5 in MLX

v0.8.4

• Added OpenAI-compatible API endpoints (/v1/chat/completions, /v1/models)
• Fixed connection issues with PC clients
• Improved server stability and error handling

v0.8.1

• Updated llama.rn adapter and updated the parameters.
• Faster chat performace - UI is optimized.
• Improved storage manager - faster model read & write performance.
• Model initialization runtime settings (context window, batch size, GPU offloading etc.) support for GGUF models.
• Branching support for chat.
• Math rendering support in chat.
• API cloning support for remote models (DeepSeek support has now been dropped).
• Fixed random app crashes.
• Many other improvements & bug fixes.

Others (Available on App Store):

• MLX support for iOS

v0.7.8

• Updated underlying llama.cpp engine

v0.7.7

• Rebranded Inferra to InferrLM
• Updated llama.rn - underlying llama.cpp engine

v0.7.5

• Update model list
• Remove portrait mode restriction
• Bug fixes

v0.7.2

• Editable base URL for remote models
• Bug fixes

v0.7.0

• GitHub re-release