| Mobile Web

Search results for DeepSeek R1

AI & Enterprise

Four teams in Korea\'s sovereign AI foundation model project ready for second review with different scale-up strategies

Four teams in the government-backed \"sovereign AI foundation model\" project have completed preparations for a second-round review, industry sources said. Two large companies have built models with more than 600 billion total parameters, comparable in size to major global open-weight models. Two startups opted for models of around 300 billion parameters, focusing on compute efficiency and task execution. All four use a mixture-of-experts structure and are pursuing different approaches to scaling up.

AI & Enterprise

China AI narrows performance gap, cuts inference costs to as little as one-hundredth

Chinese AI models are rapidly narrowing performance gaps with leading Western models while cutting inference costs to as little as one-hundredth, an analysis said. The focus is shifting from top benchmark scores to who can deploy high-performance AI cheaply and at scale. Benchmark results still show U.S. firms ahead, but Chinese models are close. Cost and open-weight strategies are boosting adoption, though political constraints and bias concerns persist.

AI & Enterprise

Perplexity cuts reasoning costs with Chinese open-source model, claims Claude Opus-level performance at one-third the cost

Perplexity released a research preview of a model it post-trained from Chinese open-source AI model GLM 5.2 for its agent system. It said the model delivers performance comparable to Anthropic\'s Claude Opus 4.8 at about one-third of the cost, routing most tasks to GLM 5.2 and sending only difficult ones to Opus 4.8. The company said the approach reduced costs versus using Opus for all tasks.