不能做什么 ❌
1. ❌ 本地跑大语言模型
A1的算力只有6 TOPS——这是给轻量任务设计的，远远不够跑大语言模型。
以下事情，A1做不到： - ❌ 本地跑DeepSeek-V3 - ❌ 本地跑Llama-3-8B - ❌ 本地跑Qwen2.5-7B - ❌ 本地跑任何7B及以上模型
如果你需要本地跑大模型，了解铠盒E1及以上的大模型计算机系列。
2. ❌ 离线使用AI
A1需要联网——因为大模型在云端。没有网，AI就不能回答复杂问题。
（本地的轻量任务不需要联网，但那些不是"聊天AI"）
3. ❌ 替代高性能电脑
A1是ARM芯片，4GB内存，不是一台通用电脑。你不能在A1上： - 打游戏 - 跑渲染 - 当办公电脑用
A1是专门的AI设备，不是通用计算机。
4. ❌ 多用户同时高强度使用
A1是家用入门款，适合个人或小家庭使用。如果一个团队同时高强度调用，性能可能不够。
适合什么人
| 适合 | 为什么 |
|---|---|
| 想用OpenClaw但不会装的人 | 省去所有安装步骤 |
| 个体户/小商家 | 微信客服自动化，成本极低 |
| 自媒体内容创作者 | AI辅助写作，微信里随时用 |
| 学生/研究者 | 文献翻译、论文润色、资料整理 |
| AI尝鲜用户 | 最低成本体验AI智能体系统 |
不适合什么人
| 不适合 | 原因 | 建议 |
|---|---|---|
| 需要本地跑大模型的人 | A1做不了 | 看看E1及以上 |
| 需要离线AI的人 | A1依赖云端 | 看看E1及以上 |
| 不想付API费用的人 | A1需要API Key | 用有免费额度的服务（DeepSeek/Gemini） |
| 需要高并发AI服务的人 | A1性能不够 | 看看F1/G1 |
| 想要一台能替代电脑的设备 | A1不是通用电脑 | 看看E1（x86，可装Windows） |
常见误解澄清
误解1："A1就是本地电脑版ChatGPT"
不是。ChatGPT是OpenAI的大模型产品，A1是运行OpenClaw的硬件。你可以用A1接入ChatGPT的API——但A1本身不是ChatGPT。
误解2："买了A1就不用付任何AI费用了"
不准确。A1是一次性买断硬件，但大模型API费用需要另外支付（取决于你选择的服务商和用量）。好消息是DeepSeek、Gemini等都有充足的免费额度。
误解3："A1能当路由器/软路由用"
不建议。A1有两个网口（物理优势），但定位是AI设备，不是路由器系统。
误解4："A1跑OpenClaw，OpenClaw是ChatGPT的开源替代"
不是。OpenClaw是智能体系统，ChatGPT是大模型产品。OpenClaw可以接入ChatGPT的API——它们是互补关系，不是竞争关系。
最后——要不要买？
买A1，如果你： - 想要一个不折腾的OpenClaw设备 - 需要微信/AI客服自动化 - 接受云端API + 本地智能体的模式
不买A1，如果你： - 需要本地跑大模型（看E1+） - 需要离线AI（看E1+） - 需要高性能通用电脑（看E1+）
延伸阅读
What the A1 Can and Cannot Do — Read Before You Buy
"Can it run models locally? Can it replace ChatGPT? Can I use it offline?"
This article doesn't sell. It tells you the truth. Know the facts before deciding.
What the A1 Is (One Sentence)
The A1 is an Agent Computer with OpenClaw pre-installed. Its design goal: let non-technical users start using AI right out of the box.
It is not a local LLM server. The "brain" comes from the cloud.
What It Can Do ✅
1. Zero-Barrier OpenClaw Access
- No Node.js installation
- No environment variable configuration
- No antivirus disabling
- No path issues
- Open browser → enter URL → start using
2. Cloud LLM API Integration
The A1 connects to these model providers:
| Provider | Models | Cost |
|---|---|---|
| OpenAI | GPT-4o, GPT-4.1 | Pay-per-use |
| Anthropic | Claude Sonnet 4, Opus 4 | Pay-per-use |
| DeepSeek | DeepSeek-V3, R1 | Very low cost + free tier |
| Gemini series | Generous free tier |
Enter your API keys, and OpenClaw taps into these models.
3. Connect to Discord/Slack/WeChat
Bring your AI assistant into your everyday messaging: - Draft content, translate, research — all from chat - Auto-reply to common customer questions - Keyword monitoring and alerts - Scheduled push notifications
4. 24/7 Online, Ultra-Low Power
- 5-12W power (under $1/month electricity)
- Fully passive cooling, silent
- Palm-sized, minimal footprint
- Designed to stay on continuously
5. Lightweight Local AI Tasks
The built-in 6 TOPS NPU handles: - Keyword extraction - Basic text classification - Sensitive information filtering
These are supplementary features, not the A1's core value.
What It Cannot Do ❌
1. ❌ Run Large Language Models Locally
The A1's 6 TOPS NPU is designed for lightweight tasks — far from sufficient for large language models.
The A1 cannot: - ❌ Run DeepSeek-V3 locally - ❌ Run Llama-3-8B locally - ❌ Run Qwen2.5-7B locally - ❌ Run any 7B+ model locally
If you need local LLMs, explore the KAIHE E1 and above — our Large Model Computer series.
2. ❌ Use AI Offline
The A1 needs an internet connection — because the LLMs are in the cloud. Without internet, the AI cannot answer complex questions.
(Lightweight local tasks don't need internet, but those aren't conversational AI.)
3. ❌ Replace a High-Performance PC
The A1 is an ARM chip with 4GB RAM — it is not a general-purpose computer. You cannot use the A1 for: - Gaming - Rendering - Desktop productivity
The A1 is a dedicated AI device, not a general-purpose computer.
4. ❌ Handle Multi-User High-Intensity Workloads
The A1 is an entry-level home device designed for personal or small-family use. Heavy simultaneous usage by a team may exceed its capacity.
Who It's For
| Ideal User | Why |
|---|---|
| OpenClaw beginners | Bypasses all installation steps |
| Small business owners | WeChat support automation, minimal cost |
| Content creators | AI-assisted writing, available in messaging apps |
| Students/researchers | Translation, paper polishing, note organization |
| AI explorers | Lowest-cost entry into AI agent systems |
Who It's Not For
| Not Suitable | Reason | Alternative |
|---|---|---|
| Local LLM users | A1 can't do it | Explore E1+ |
| Offline AI users | A1 depends on cloud | Explore E1+ |
| Users unwilling to pay API fees | A1 needs API keys | Use free-tier providers (DeepSeek, Gemini) |
| High-concurrency AI needs | A1 performance ceiling | Explore F1/G1 |
| Users wanting a PC replacement | A1 is not a general-purpose computer | Explore E1 (x86, runs Windows) |
Common Misconceptions
Misconception 1: "A1 is a local version of ChatGPT"
No. ChatGPT is OpenAI's LLM product. The A1 is hardware running OpenClaw. You can use the A1 to access ChatGPT's API — but the A1 itself is not ChatGPT.
Misconception 2: "Once you buy the A1, there are no more AI costs"
Inaccurate. The A1 hardware is a one-time purchase, but LLM API costs are separate and depend on your chosen provider and usage. Good news: DeepSeek, Gemini, and others offer generous free tiers.
Misconception 3: "The A1 doubles as a router"
Not recommended. While the A1 has two Ethernet ports, it's designed as an AI device, not router hardware.
Misconception 4: "OpenClaw is an open-source replacement for ChatGPT"
No. OpenClaw is an agent system; ChatGPT is a language model product. OpenClaw can access ChatGPT's API — they are complementary, not competitive.
The Bottom Line
Buy the A1 if you: - Want a zero-hassle OpenClaw device - Need messaging-platform AI automation - Accept the cloud API + local agent model
Don't buy the A1 if you: - Need local LLM inference (explore E1+) - Need offline AI (explore E1+) - Need a high-performance general-purpose computer (explore E1+)