Chinese Open-Weight AI Models Overtake U.S. Counterparts in Downloads and Benchmarks
सिर्फ तथ्य

Chinese Open-Weight AI Models Overtake U.S. Counterparts in Downloads and Benchmarks

सारांश

Chinese companies have surpassed U.S. firms in the adoption and performance of open-weight language models, according to recent congressional briefing data.

Open-weight language models—AI systems whose weights are publicly available—are now a major focus of U.S.-China competition. While closed models such as GPT-4 are accessed only via APIs, open-weight models like Meta's Llama, Alibaba's Qwen and Google's Gemma can be downloaded and run locally.

Data tracked by the briefing author shows that Chinese open-weight models have led U.S. models in Hugging Face downloads since July 2025, reaching roughly 1.6 billion downloads out of a total of 3.2 billion. Benchmark scores on the Artificial Analysis Intelligence Index also favor Chinese models; as of September 14, 2026, Z.ai’s GLM-5.3, GLM-5.3-Flash and Moonshot AI’s Kimi K3 scored 45, 42 and 44, compared with 26 for the leading U.S. models Inkling and Inkling Small.

"Chinese labs release models faster and focus on a narrower set of tasks, which helps them achieve higher benchmark scores," the briefing notes.

The gap between open-weight and closed models has narrowed to an estimated two-to-five-month difference in capabilities, with Chinese open-weight models typically a few months ahead of U.S. open-weight offerings and six-to-nine months behind leading closed models from OpenAI and Anthropic.

Adoption of Chinese open-weight models is growing across platforms that route inference requests, such as OpenRouter, where Chinese models now account for over 80 % of token usage. Major U.S. companies—including legal, coding and food-delivery services—have incorporated these models into their products, and academic research on arXiv increasingly cites Chinese models, which appear in more than 40 % of relevant papers.

The briefing highlights regulatory uncertainty surrounding the open release of powerful models, noting that restricting access could inadvertently harm U.S. businesses that rely on these tools. It recommends continued U.S. investment in open-weight AI to maintain strategic and economic advantages while preparing for emerging security risks.

स्रोत

Interconnects AI

पहली बार यहां रिपोर्ट किया गया

पढ़ते रहें — निःशुल्क

समाचार फ़ॉलो करने के लिए मुफ़्त खाता बनाएँ। कार्ड की आवश्यकता नहीं।