TeacherPro v2.3.0

TeacherPro v2.3.0 – Fresh AI Model Catalog

Changed

🧠 Updated local AI model catalog

The downloadable model list now features current generations instead of older
2024/early-2025 models. Every entry runs locally via Ollama and is sized for
real hardware — from 8 GB RAM laptops up to a 24 GB VRAM GPU (RTX 4090 class).

Small (8 GB RAM)

• Granite 4.2 3B
• Qwen 3.5 4B
• Granite 4.2 8B
• LFM 2.5 8B
• Ministral 3 8B

Medium (16 GB RAM / 12 GB VRAM)

• Qwen 3.5 9B — still the recommended default
• Gemma 4 E2B
• Gemma 4 12B
• Ministral 3 14B
• Gemma 4 E4B
• GPT-OSS 20B

Large (24 GB VRAM)

• Qwen 3.6 27B
• Qwen 3.8 27B
• Gemma 4 26B
• Gemma 4 31B
• Qwen 3.6 35B

Thinking-mode support was extended to the new Qwen 3.6/3.8, Gemma 4 12B/31B and
GPT-OSS models.

🗑️ Removed outdated models

Llama 3.1 8B, Llama 3.2 3B, DeepSeek R1 8B/14B, Mistral Nemo 12B,
Mistral Small 3.1 24B and Qwen 3.5 27B have been removed in favour of the
current models above. The bge-m3 embedding model is unchanged.

Removed

• Unused legacy modelCatalog translation strings in src/i18n/de.ts and en.ts.

Files Changed

• src/ai/modelCatalog.ts – refreshed model catalog and thinking-model list
• src/i18n/de.ts, src/i18n/en.ts – removed unused model catalog strings
• CHANGELOG.md – v2.3.0 entry