I’ve been playing around with the Multi-Model feature in Anna lately, and I’ve realized something interesting: choosing a model isn’t just about the “intelligence” score or the price—it’s about the personality and style.
Here are my personal findings:
The Literary Champion: Gemini 3 Flash Preview
I think this is a hidden gem in Anna’s model pool. Its literary level is outstanding—the way it crafts sentences feels much more “human” and creative than others. Plus, at only x6 multiplier, the value for money is unbeatable if you are doing creative writing or storytelling.
The “Cold” Logic: GPT-4
GPT remains powerful, but man, it feels cold. It’s direct, efficient, and professional, but sometimes it feels like talking to a very smart but very bored professor. Perfect for facts, but not for inspiration.
The “Warm” Neighbor: Qwen
On the complete opposite side, Qwen is incredibly warm. Sometimes it’s even a bit too chatty (talks a lot! lol), but its tone is so encouraging and gentle. It feels like that one neighbor who always wants to make sure you’re doing okay.
It’s fascinating how Anna lets us switch between these “brains” depending on our mood or the task at hand. Has anyone else noticed these personality traits in other models?
This is such a fantastic breakdown! I love your descriptions – especially the “bored professor” for GPT-4 and the “chatty neighbor” for Qwen. Those are spot-on!
You have nailed something really important: model selection is deeply personal. It is not just about benchmarks or pricing – it is about finding the “voice” that matches your workflow and mood.
My Personal Favorite: Qwen3.5 Plus
I have been using Qwen3.5 Plus as my daily driver, and I think it is the king of cost-performance! For most day-to-day tasks, it strikes this perfect balance where I can use it long-term without worrying about burning through credits. It has become my “default brain” – reliable, warm, and efficient.
Quick fact-checking: Any fast model (no need to overthink it)
Long research sessions: Qwen3.5 Plus (budget-friendly for extended use)
Image Tasks:
Image editing: Qwen Image Edit Max – it understands context really well and makes precise adjustments
Image generation: Doubao Seedream 5.0 – the quality is outstanding for creative work, especially for artistic styles
It is awesome that Anna gives us this flexibility. Having the right tool for the right job makes such a difference!
Question for You: Have you tried any of the Claude models? I would love to hear how they compare to your “literary champion” Gemini for writing tasks! Also, do you use the image generation/editing features? If so, which models do you prefer?
Thanks for sharing this – it is a great guide for other users figuring out their model preferences!
I love your descriptions – especially the “bored professor” for GPT-4 and the “chatty neighbor” for Qwen. Those are spot-on! 
Text Tasks:
Image Tasks: