Alibaba says its new AI model can read your emotions

Kelechi Edeh -

AI has gotten scary good at understanding what we type, say, and even mean—but what if it could actually see how we feel?

That’s exactly what Chinese tech giant Alibaba is trying to do with its latest open-source AI model, R1-Omni. Unlike traditional AI models that just analyze text, this one watches you—tracking facial expressions, body language, and even environmental context to infer emotions. In a demo, Alibaba showed R1-Omni identifying emotions from video footage while also describing what people were wearing and where they were. So yeah, it’s basically computer vision meets emotional intelligence.

And while emotion-detecting AI isn’t entirely new (Tesla already uses AI to spot drowsy drivers), Alibaba’s model could take things a step further—putting emotion recognition in an open-source package that anyone can download for free.

Alibaba’s new AI model to rival that of DeepSeek and OpenAI
But can the tech giant offer a compelling alternative that balances cost and performance?
TechloyKelechi Edeh

Also, the timing of this launch may not be a coincidence. Just last month, OpenAI dropped GPT-4.5, touting better emotional nuance detection in conversations. But there’s a key difference—GPT-4.5 is strictly text-based, meaning it can infer emotions from what you write, but it can’t actually see them. There’s also the pricing factor: GPT-4.5 is locked behind a paywall ($20/month for Plus, $200/month for Pro), while Alibaba’s R1-Omni is completely free on Hugging Face.

But this isn’t just about one-upping OpenAI. Alibaba has been on a full-blown AI rampage ever since DeepSeek (another Chinese AI startup) shook up the industry by outperforming ChatGPT in some benchmarks. Now, every major Chinese tech giant is racing to keep up—and Alibaba is leading the charge. The company has been benchmarking its Qwen model against DeepSeek, partnering with Apple to bring AI to iPhones in China, and now throwing emotion-aware AI into the mix to keep OpenAI on its toes.

Of course, R1-Omni isn’t a mind reader (yet). It can recognize emotions, but it doesn’t exactly react to them– at least for now. But let’s be real, if AI can already tell when we’re happy or annoyed, how long before it starts tailoring its responses based on our moods?

Now that is a little unsettling.