AnswersPublished17 sources
China’s Top AI Models Are Now About 3% Behind US Leaders
Bloomberg Intelligence estimates that leading Chinese AI models trail US rivals by about 3% on benchmarks, down from roughly 15% earlier in 2026. DeepSeek V4.1 Flash scored 81.1 on LiveBench and ranked sixth globally, compared with 83.4 for Anthropic’s leading model.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Bloomberg Intelligence estimates that leading Chinese AI models trail US rivals by about 3% on benchmarks, down from roughly 15% earlier in 2026.
- DeepSeek V4.1 Flash scored 81.1 on LiveBench and ranked sixth globally, compared with 83.4 for Anthropic’s leading model.
- The shrinking gap raises questions about US chip export controls and AI leadership, while Chinese firms still face limits on resources and challenges turning competitive models into sustainable profits.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: How does Bloomberg Intelligence assess the narrowing performance gap between leading US and Chinese AI models in 2026, including DeepSeek V4. Article summary: Bloomberg Intelligence (BI) says the best Chinese AI models have moved much closer to US leaders in 2026: their benchmark-performance gap fell to about **3%** after DeepSeek’s September release, from roughly **9% in May*. Topic tags: general, news, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers
Bloomberg Intelligence (BI) says the performance gap between leading Chinese and US AI models narrowed to about 3% after DeepSeek released V4.1 Flash in September 2026. BI had put the gap at roughly 9% in May and 15% earlier in the year. The figures point to a fast-moving contest in model performance—not a verdict that China has surpassed the US across the broader AI industry.
What DeepSeek V4.1 Flash’s LiveBench score shows
DeepSeek V4.1 Flash scored 81.1 on LiveBench, placing sixth in the global ranking cited in coverage of BI’s assessment. Anthropic’s leading model scored 83.4. The 2.3-point difference between those scores is consistent with BI’s estimate of an approximately 3% performance gap. 12
That comparison is a result on one benchmark, not a complete measure of AI capability or the entire US–China technology race. It shows how close these particular leading models scored on LiveBench; it cannot establish that every model, task or part of the AI business is equally close.
A narrowing gap, but not yet a broad Chinese lead
The change has developed over several months. BI reported that Chinese models had narrowed the gap to 6% in June, down from 9% in May, and had secured two spots in LiveBench’s global rankings at that time. The later estimate of about 3% suggests the pace of improvement continued. 2
The rankings also show why “closing in” is more accurate than “overtaking.” In the later LiveBench ranking reported alongside the new estimate, Chinese models occupied only three of the top 15 positions. A narrow gap between the top-scoring models does not mean China has the same depth of high-ranked models. 8
What the results mean for chip controls and US AI leadership
Chinese labs’ progress while facing restrictions on access to advanced chips raises questions about how effectively export controls can preserve a US lead in model performance. But the benchmark results alone do not show that the restrictions have had no effect. BI has also noted that Chinese AI companies have less financial firepower than US rivals and lack access to the most cutting-edge chips. 1
17
The careful conclusion is that the US lead on this measure has narrowed sharply. Whether that changes the broader balance of AI leadership depends on more than a single leaderboard, including the resources available to companies and their ability to turn models into widely used, profitable products.
Why strong models do not guarantee strong profits
Model quality is only part of the business equation. Bloomberg’s reporting describes Chinese models as a competitive low-cost alternative, while also noting monetization barriers for the sector. DeepSeek’s steep pricing can increase competitive pressure, but attracting users with low prices does not by itself demonstrate that a company can earn sustainable profits. 3
18
Bottom line: BI’s assessment marks a significant narrowing in the measured performance gap: from about 15% earlier in 2026 to about 3% after DeepSeek V4.1 Flash. The result challenges assumptions about a lasting US advantage in model performance, but it is not proof of a comprehensive Chinese lead—and the path from competitive models to durable profits remains uncertain. 3
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "China’s Top AI Models Are Now About 3% Behind US Leaders"?
Bloomberg Intelligence estimates that leading Chinese AI models trail US rivals by about 3% on benchmarks, down from roughly 15% earlier in 2026.
What are the key points to validate first?
Bloomberg Intelligence estimates that leading Chinese AI models trail US rivals by about 3% on benchmarks, down from roughly 15% earlier in 2026. DeepSeek V4.1 Flash scored 81.1 on LiveBench and ranked sixth globally, compared with 83.4 for Anthropic’s leading model.
What should I do next in practice?
The shrinking gap raises questions about US chip export controls and AI leadership, while Chinese firms still face limits on resources and challenges turning competitive models into sustainable profits.
Sources
aiweekly.coBloomberg Intelligence: US Lead Over China in Top AI Models Narrows to Record 3% After DeepSeek V4.1 Flash | AI Weekly
bloomberg.comChina Cuts AI Gap as Moonshot Shows Zhipu Isn't One-Off ...
mitrade.comIs America Losing Its AI Lead to China? The Gap Is Now 3%