The quiet spell is officially over. This week, a new class of “vision” models hit Ollama hard, while Google and DeepSeek finally dropped the hardware-gap-fillers we’ve been waiting for.

This isn’t a maintenance patch; it’s a major arsenal update.

Here’s the latest intel on our almanac:

🔹Before: 83 models
🔹New Drops: 4
🔸New Total: 87 models

Let’s break down the new powerhouses you can run on your own hardware right now.

👁️ The Vision Beasts: Qwen3-VL (8B & 32B) by Alibaba

LLaVA and Phi-3 Vision were great, but this is the contender we were waiting for. Qwen has always been a monster in multimodal, and their latest vision-language models just landed in Ollama. We’ve added two flavors: a lightweight 8B for daily use and a 32B beast for heavy-duty analysis on 24GB cards.

Qwen3-VL 8B:

• VRAM: ~8GB
• Size: 4.9GB
• Ollama Name: qwen3-vl:8b

Qwen3-VL 32B:

• VRAM: ~24GB
• Size: 19.5GB
• Ollama Name: qwen3-vl:32b

🧩 The 12GB Gap-Filler: Gemma 3 12B by Google

We all felt this gap. You had 10GB models (like Phi-4) and then a jarring jump to 16GB+. If you were running a 12GB card, you were left out. Google just plugged that hole with their third-gen Gemma, built specifically for solid reasoning in that midrange sweet spot.

• VRAM: ~12GB
• Size: 7.0GB
• Ollama Name: gemma3:12b

🧠 The Logic Machine: DeepSeek-R1 14B by DeepSeek

Same logic, different beast. DeepSeek continues its reign in coding and reasoning, and this 14B model is the perfect balance of power and efficiency. It’s the ideal choice for 16GB cards, leaving you power to spare without sacrificing top-tier performance.

• VRAM: ~14GB
• Size: 8.1GB
• Ollama Name: deepseek-r1:14b

Ready to Run the Future?

The era of simple text-bots is ending. The next wave of local AI is here—models that can see and reason just as well as they chat.

You don’t need to lift a finger.

Just launch LLM VibeCheck. The app syncs the new almanac automatically. Get your hands on the next generation of AI right now.

 

Follow the Vibe