v0.12.3
📦 ollamaView on GitHub →
✨ 3 features🐛 3 fixes🔧 5 symbols
Summary
This release adds support for DeepSeek-V3.1-Terminus and Kimi-K2-Instruct-0905 models while fixing critical bugs related to tool call parsing, Unicode rendering, and model loading crashes.
✨ New Features
- Added support for DeepSeek-V3.1-Terminus, a hybrid model supporting both thinking and non-thinking modes.
- Added support for Kimi-K2-Instruct-0905, a 1-trillion parameter mixture-of-experts (MoE) model.
- ollama push now provides a sign-in URL for authentication.
🐛 Bug Fixes
- Fixed issue where tool calls provided as stringified JSON were not parsed correctly.
- Fixed incorrect Unicode character output for qwen3-coder models.
- Fixed a crash occurring when loading models via the /load endpoint.
🔧 Affected Symbols
/loadollama pushdeepseek-v3.1kimi-k2qwen3-coder