AI tool comparison
QwenPaw vs Sup AI
Which one should you ship with? Here is the side-by-side panel verdict, pricing read, reviewer split, and community vote comparison.
Personal AI
QwenPaw
Self-hosted personal AI with evolving memory, runs on 6+ chat apps
75%
Panel ship
—
Community
Free
Entry
QwenPaw (formerly CoPaw, rebranded April 2026) is an open-source personal AI assistant built by the AgentScope team at Alibaba. You deploy it locally or on a cloud VM, connect it to messaging apps like Telegram, Discord, WeChat, DingTalk, or Feishu, and interact with a persistent, memory-evolving agent that learns your preferences and proactively surfaces relevant information. Version 1.1.4, released April 24, brings a refactored memory and context architecture, built-in DeepSeek V4 models, ACP Server exposure for multi-agent communication, and a console plugin system. For LLM backends it supports cloud APIs (Qianwen, DeepSeek, OpenAI) and fully offline local inference via Ollama, LM Studio, or llama.cpp — meaning you can run it with zero API costs on your own hardware. The built-in skill library covers daily news digests, video summarization, email triage, PDF/Office processing, and calendar management. The multi-agent capability — where you can spin up specialized agents that collaborate — puts it in interesting territory between a personal assistant and a lightweight team-of-agents platform. Desktop apps for Windows and macOS are in beta.
AI Assistants
Sup AI
Confidence-weighted AI ensemble that topped Humanity's Last Exam
67%
Panel ship
—
Community
Free
Entry
Sup AI uses a confidence-weighted ensemble of multiple AI models to answer hard questions. Each model rates its own confidence, and the system aggregates responses weighted by that confidence. Achieved 52.15% on Humanity's Last Exam benchmark, outperforming individual models.
Reviewer scorecard
“The Ollama backend support is the key feature — this is the first personal assistant I've seen where you can genuinely go fully offline and fully free. The ACP server in v1.1.4 opens it up for multi-agent coordination that's actually useful for automating dev workflows.”
“No API, no self-hosting option, and the ensemble approach means your per-query cost is 3-5x a single model call. The benchmark numbers are compelling but I cannot integrate this into a product. Ship an API and I will reconsider.”
“The skill library looks impressive on paper but most of the demos are China-centric platforms (Xiaohongshu, Zhihu, DingTalk). International users will find meaningful gaps and will need to build their own skills. The documentation is also still primarily in Chinese despite multilingual README efforts.”
“The benchmark result is legitimately impressive and the methodology is transparent. My concern is latency — querying multiple models and aggregating adds significant time. For research and high-stakes questions it is worth the wait. For everyday chat it is overkill.”
“The future of personal AI is self-hosted, memory-persistent, and connected to where you actually communicate. QwenPaw's architecture — LLM backend agnostic, multi-platform, multi-agent — is the right shape for that future. The Alibaba team building this in the open is a meaningful contribution.”
“Confidence-weighted ensembling is the quiet breakthrough everyone is sleeping on. Individual models plateau — but smart aggregation keeps pushing the frontier. Sup AI scoring 52% on Humanity's Last Exam when no single model breaks 40% proves the thesis.”
“The 'describe your goal before sleep, wake up to a prototype' workflow is the creator feature I didn't know I needed. Video pipeline automation and newsletter digests pushed to Telegram cover 80% of my daily content research. This one's getting installed.”
Weekly AI Tool Verdicts
Get the next comparison in your inbox
New AI tools ship daily. We compare them before you waste an afternoon.