Impact Score: 5/10
Ollama v0.12.3 — Release Notes & Engineering Impact
Source available — automated, not human-reviewed
Component: ollama · Channel: stable. Automated analysis, not a compatibility certification. How this works.
Analysis updated: 2026-11-01.
⚡ Executive Summary (TL;DR)
Ollama v0.12.3. Source highlight: DeepSeek-V3.1-Terminus: DeepSeek-V3.1-Terminus is a hybrid model that supports both thinking mode and non-thinking mode. It delivers more stable & reliable outputs across benchmarks compared to the previous version:
Who Does This Impact? (Role-Based Analysis)
💻 Software Engineers & Devs
Check the source notes against the package and APIs you use before upgrading.
📊 Product & Engineering Leads
Use the documented changes to assess relevance to your product; release frequency alone is not a quality measure.
⚙️ DevOps & Security
Verify the target version, runtime requirements and rollback procedure in the official documentation.
🔧 Bug Fixes & Patches (3)
- Fixed issue where tool calls provided as stringified JSON would not be parsed correctly
- Fixed issues where qwen3-coder would output unicode characters incorrectly
- Fix issue where loading a model with `/load` would crash
📄 View Raw Release Notes (Original Developer Log)
## New models
- [DeepSeek-V3.1-Terminus](https://ollama.com/library/deepseek-v3.1): DeepSeek-V3.1-Terminus is a hybrid model that supports both thinking mode and non-thinking mode. It delivers more stable & reliable outputs across benchmarks compared to the previous version:
Run on [Ollama's cloud](https://ollama.com/cloud):
```
ollama run deepseek-v3.1:671b-cloud
```
Run locally (requires 500GB+ of VRAM)
```
ollama run deepseek-v3.1
```
- [Kimi-K2-Instruct-0905](https://ollama.com/library/kimi-k2): Kimi K2-Instruct-0905 is the latest, most capable version of Kimi K2. It is a state-of-the-art mixture-of-experts (MoE) language model, featuring 32 billion activated parameters and a total of 1 trillion parameters.
```
ollama run kimi-k2:1t-cloud
```
## What's Changed
* Fixed issue where tool calls provided as stringified JSON would not be parsed correctly
* `ollama push` will now provide a URL to follow to sign in
* Fixed issues where qwen3-coder would output unicode characters incorrectly
* Fix issue where loading a model with `/load` would crash
## New Contributors
* @gr4ceG made their first contribution in https://github.com/ollama/ollama/pull/12385
**Full Changelog**: https://github.com/ollama/ollama/compare/v0.12.2...v0.12.3