On October 7, 2026, a report said Chinese developers released 16 AI models during September, including models from DeepSeek and Xiaomi. The count captures a month of launches; a separate benchmark comparison addresses model scores.
Chinese developers reportedly launched 16 models in September
The reported total covers Chinese developers’ model releases during September 2026. DeepSeek and Xiaomi were among the companies releasing models. A release count tracks activity, not performance: the number alone does not say how a model fares on an evaluation.
What the reported benchmark comparison shows
A separate October 4, 2026 estimate put the benchmark-score gap between leading Chinese and U.S. models at 3%, down from 15% at the start of the year and 9% in May. The figure describes a reported score comparison, not a general rating of every model or capability.
The reported LiveBench results put DeepSeek Flash V4.1 at 81.1 points and sixth place globally. Anthropic Fable 5.1 led the ranking with 83.4 points. The two scores are 2.3 points apart.
| Model | Developer | LiveBench score | Reported global rank |
| DeepSeek Flash V4.1 | DeepSeek | 81.1 points | Sixth |
| Fable 5.1 | Anthropic | 83.4 points | First |
LiveBench scores reflect performance on a particular evaluation. They offer one way to compare the two named models, while the reported 3% gap concerns the broader benchmark-score comparison. For more background on open-weight models, read our earlier coverage of Mozilla’s report.
AI releases and safety calls
Anthropic CEO Dario Amodei has called for slowing AI development over safety concerns. Those calls are part of the wider debate over the pace of AI development.
