I Tried Manus AI Over The Weekend
The LLMs I’ve most recently used include Anthropic’s Claude 3.7 Sonnet Pro, Google’s Gemini 2.5 (the best LLM for coding at the moment), OpenAI’s ChatGPT (free version; by the way - OpenAI’s recent $340 billion valuation was way overvalued pre-last week’s equity market slide. Softbank overpaid as per usual.) I’ve used DeepSeek once or twice, but not at length.
In terms of AI Agents, I’ve used Replit at length and kicked the tires its “Pro” version last month.
This weekend I used Manus AI’s agent. I was more impressed by Manus than Replit. However, my Manus usage was limited to approximately 90 mins of work as Manus only allocates 1,000 “credits” to its free tier, and those credits are quickly utilized as Manus runs code and makes API calls. However, the one medium difficulty programming task that I assigned Manus, it executed well. Manus’ value add is the APIs it has built into third-party applications that allow it to provide agentic capability.
I’m not sure if I will kick the tires on Manus’ Pro tier, which starts at $39/month. Beyond the one TEK2day technology project I have in the queue, there is nothing to follow, and Anthropic is more than capable of running the small queries that I may require.
My sense is that Google (GOOG) would wipe Manus (and Replit) off the map if Google kept Gemini 2.5 free, and added APIs the way Manus has, while charging $20/month for a consumer Agentic AI offering. Another Agentic AI pricing model that Google may follow would be to offer a free tier to pull in users and only charge “power” users a usage fee beyond a certain consumption threshold.