ByteDance Challenges OpenAI and Google with Powerful, Affordable AI

ByteDance’s Doubao-1.5 Pro surpasses GPT-4 and Claude 3.5 in performance while cutting costs by up to 70%, putting pressure on artificial intelligence giants

Let’s talk about the new competitor in the world of AI that is making giants like OpenAI and Google sweat. This week, ByteDance — the company behind TikTok — released Doubao-1.5 Pro, a large-scale language model (LLM) that not only competes with GPT-4o and Claude 3.5 Sonnet, but outperforms both in key areas. Most impressively, he does this while reducing costs and maintaining his “no shortcuts” philosophy.

But before we dive into the details, here’s a tip for developers: want to test the Doubao-1.5 Pro API? Download Apidog for free and start integrating this powerful tool into your projects today. Trust me, you’ll want to try it.

What makes the Doubao-1.5 Pro a game-changer?

1. Performance that surpasses the giants
   Let’s get straight to the point: the Doubao-1.5 Pro isn’t just good — it rocks the benchmarks. ByteDance claims it outperforms heavyweights like GPT-4o and Claude 3.5 Sonnet in areas like knowledge retention, programming, reasoning, and Chinese language processing. As? Using a sparse Mixture of Experts (MoE) architecture, which uses fewer activation parameters during training, but delivers the performance of a dense model seven times higher. Translation: it’s like having the performance of a Lamborghini for the cost of a Toyota.

And there’s more: the Doubao-1.5 Pro didn’t just surpass its predecessors — it left them behind. Compared to the Skylark model, also from ByteDance, it shows a 19% increase in performance in math, language comprehension and multitasking assessments.

2. Attention-grabbing cost efficiency
   Let’s go to the numbers. The Doubao-1.5 Pro-32k model costs just ¥0.8 (about $0.11) per million tokens for input, with output tokens costing ¥2. Even the lightest version, Doubao-1.5-lite-32k, reduces prices even further, to ¥0.3 per million input tokens. For context, this is about 70% cheaper than the industry average.

How did ByteDance achieve this? Two words: operational efficiency. By optimizing server clusters and network card technology, they have reduced hardware costs while increasing efficiency in small packet communication. The result? An impressive 50% gross margin for the Pro Model API — a number most AI companies only dream of achieving.

Why are developers talking so much about Doubao-1.5 Pro?

1. Easy integration (no PhD needed)
   ByteDance knows that even the best AI is useless if it is difficult to implement. So they simplified the deployment process:

• Sign up to Volcano Engine (your cloud platform) or use an existing account.
• If you don’t speak Chinese, Google Translate will be your best friend. Right-click on the page and select “Translate to English” (Chrome does this automatically). For more complicated sections, take a screenshot and upload it to the Google Translate app.

Still unsure? Learn these quick phrases:

• API 密钥 (API mì yuè) = Chave da API
• Model Selection (Móxíng xuǎnzé) = Seleção de modelo
• Register (Zhùcè) = Cadastro

Choose your model.

Click here to start testing.

From this interface, you can ask anything you want.



2. Independence in training: without depending on external data
   Unlike some models that rely on external data, Doubao-1.5 Pro is trained on ByteDance’s completely independent data ecosystem — think books, forums, news, and more. This not only ensures compliance but also reduces bias, making the model more reliable for business use.

The Big Picture: What This Means for the Future of AI

ByteDance isn’t just after profits — they’re on a mission to make AI accessible. As Tan Dai, president of Volcano Engine, said: “Price cuts should not be limited to light models. Core models need to remain affordable to drive innovation.” By keeping prices low even for high-end models, they are inviting startups and independent developers to experiment where they couldn’t before.

Of course, not everything is rosy. ByteDance’s AI division is not yet profitable due to huge investments in R&D. But with adoption growing exponentially (API calls on smart endpoints increased 100x in six months), economies of scale could be a game changer.

For businesses, the lesson is clear: now is the time to explore Doubao-1.5 Pro. Whether building a chatbot, analyzing data, or creating multimodal content, this model offers a rare combination of power and accessibility.