TL;DR
Jack Roberts demonstrates how to connect MiniMax M3 to the Hermes AI agent as an alternative to Claude and ChatGPT, showing it matches GPT-4.5 on coding benchmarks at roughly 1/25th the price. The core idea is being model-agnostic and routing each task to the best, most cost-efficient model rather than defaulting to expensive flagship models.
Key points
1
MiniMax M3 matches GPT-4.5 on coding benchmarks at about 4% of the price, and edges Claude Sonnet 4.6 on some tasks at 8% of the price
2
Hermes agent acts as the car (framework), while AI models are swappable engines — you build the agent once and drop in different models per task
3
MiniMax M3 achieves its efficiency via sparse attention, which reduces compute by roughly 20x by only processing contextually relevant tokens
4
Connecting MiniMax M3 to Hermes via Telegram takes just an API key — select Minimax Global in terminal, paste the key, then switch models inside Telegram
5
Hermes can dynamically route queries to different models automatically using a routing skill, letting you pre-allocate specific models to specific task types
Actionable insights
→
Go to platform.minimax.io, grab an API key, and connect it to Hermes by selecting Minimax Global in terminal — the $20 plan gives 1.7 billion tokens
→
Use the most powerful model (e.g. Claude Opus or GPT-5) for initial planning and strategy, then delegate execution subtasks to cheaper models like MiniMax M3 to cut costs at scale
→
Build a dynamic routing skill inside Hermes so it automatically selects the right model for each task — Jack has a pre-built skill available in the video description
→
MiniMax M3 supports 1 million token context, native image and video understanding, and live web browsing — use it specifically for multimodal and web-scraping tasks where other models charge much more
Notable quotes
“You own the car and swap the engine.”
“Imagine a model that is as capable as GPT 5.5, but is only like 1/25th of the price.”
“You own the agent and rent the brain.”
Worth watching?
⏭️
Worth watching the full video?
The key setup steps, pricing comparisons, and strategy are all captured here — only watch the full video if you want to see the live Telegram demo and Firecrawl scraping walkthrough in action.
Saved you some time? The creator still deserves a like.
Found this by searching? That's the hard way.
summree watches the AI channels you follow and sends you the summary the moment a video drops. Nothing gets past you. From £4/month.
Try it free