Skip to content

Vesta 0.9.6

Choose a tag to compare

View all tags
@scouzi1966 scouzi1966 released this 04 Mar 14:15
· 4 commits to main since this release
v0.9.6
3b6402d

Vesta 0.9.6

Download: Vesta-0.9.6.dmg

SHA256: 0f4acd25c4b042d57e4e8716e04d3cf993e28bd149c47f7136a3646546cfcf09

Install via Homebrew:

brew tap scouzi1966/afm
brew install --cask scouzi1966/afm/vesta-mac

Changes since v0.9.5:

  • Add Qwen3.5 reasoning model support (enable_thinking toggle for MLX)
  • Add qwen3_5, qwen3_next, GLM-5, GLM-4.7-Flash, Kimi K2, MiniMax-M2, Nemotron-H architecture support
  • Add implicit reasoning for llama.cpp and MCP prompt guardrails
  • Add modality pills with load/unload lifecycle management
  • Add Sidekick save_file MCP tool with inline file cards
  • Add OpenWhisper STT routing with MLX parity
  • Upgrade llama.cpp b7847→b8182 with Metal performance improvements
  • Add security-scoped bookmarks for imported GGUF models
  • Fix MLX memory leak on model unload and stale model name
  • Fix TTS duplicate messages, stale DerivedData builds
  • Comprehensive test infrastructure: unit, UI, MCP contract, E2E pipeline
Assets 3
Loading