Baidu (BIDU) Stock; Climbs as Ernie 5.1 Slashes AI Training Costs

TLDRs;

• Baidu stock rose slightly after launching Ernie 5.1, claiming a 94% reduction in AI training costs.
• The new model reportedly matches top global AI systems while using far less computing power.
• Ernie 5.1 ranks highly on benchmarks, behind only leading US-based AI models.
• The release highlights China’s push for efficient AI amid US chip export restrictions.

Baidu shares edged slightly higher after the company unveiled its latest artificial intelligence model, Ernie 5.1, which it claims dramatically reduces the cost of training advanced AI systems.

The Chinese tech giant says the model delivers performance comparable to leading global AI systems while requiring only about 6% of the pre-training computing cost typically needed. The announcement drew immediate attention from investors, highlighting Baidu’s push to strengthen its position in the increasingly competitive global AI landscape.

New Efficiency Model Redefines Training Approach

At the heart of Ernie 5.1 is a major shift in how AI models are trained. Baidu attributes the cost reduction to a more efficient architecture and its “Once-For-All” training framework, which is designed to reduce dependence on massive GPU clusters. Instead of training a single large model from scratch in a resource-heavy process, the system trains multiple smaller configurations simultaneously within one unified cycle.

Baidu, Inc., BIDU

The final model is then extracted as an optimized sub-network from a larger structure, significantly reducing redundant computation. This approach marks a strategic shift away from traditional brute-force scaling methods toward a more resource-efficient design philosophy. According to Baidu, this is what enables the substantial reduction in computing requirements compared with earlier versions of its AI systems.

Strong Benchmark Performance Despite Skepticism

Despite questions from some industry observers regarding the validity of its cost-saving claims, Ernie 5.1 has posted strong results on key performance benchmarks. On the Arena evaluation platform, the model ranked first among Chinese AI systems and fourth globally. It trailed only advanced models developed by OpenAI and Anthropic, suggesting that Baidu is closing the gap with leading Western AI developers while simultaneously lowering costs.

However, not everyone in the AI industry is convinced. Some researchers and analysts have expressed skepticism about whether the reported cost reductions fully reflect all stages of training and infrastructure usage. They argue that real-world deployment costs and hidden compute requirements could paint a more complex picture than the headline figures suggest. Still, the benchmark performance has been strong enough to maintain investor interest.

Global AI Race Shifts Toward Efficiency

The launch of Ernie 5.1 comes at a time when Chinese technology firms are increasingly focusing on efficiency due to tightening US export restrictions on advanced semiconductor technology. Limited access to high-end GPUs has forced companies like Baidu to innovate under constraint, prioritizing models that can achieve strong performance with fewer computing resources.

This broader environment has accelerated a shift in AI development strategy. Instead of relying purely on scaling up model size and infrastructure, firms are now exploring more efficient architectures that optimize performance per unit of compute. Baidu’s approach reflects this trend and highlights how geopolitical and supply chain pressures are shaping technological innovation.

The post Baidu (BIDU) Stock; Climbs as Ernie 5.1 Slashes AI Training Costs appeared first on CoinCentral.