A new way of measuring AI models suggests China's threat to OpenAI and Anthropic may have been overblown
A revised ranking by Artificial Analysis reveals a larger performance gap between leading US AI models and Chinese open-weight models. Under the updated system—with secret tests to prevent benchmark hacking—the best open-weight models scored 32 points, while Anthropic’s Claude Opus 5.5 led with 58. Even Anthropic’s smaller model, Claude Haiku 5.5, outperformed all open-weight models in complex…
Key points
- **DeepSeek** traffic on OpenRouter could run on a $6M–$7M Nvidia server, per OpenCode’s estimate
- Open-weight models still excel in software development but face broader performance gaps in complex tasks
The shift challenges concerns that cheaper, freely available Chinese AI models could undermine US firms like OpenAI and Anthropic. Dax Raad of OpenCode found that traffic for DeepSeek—a popular open-weight model—through OpenRouter could be handled by a single $6M–$7M Nvidia server, suggesting lower real-world demand than industry hype implies. Nathan Lambert of Trillium Labs acknowledged the previous benchmark may have overstated open-weight models’ strength, though they remain competitive in software development and offer cost advantages. The findings could bolster investor confidence in Anthropic’s IPO and OpenAI’s potential listing, reinforcing the narrative of a sustained US lead.
Model pages: Claude Opus 5.5 → · Claude Haiku 5.5 → · Kimi K3.1 Pro →
A new way of measuring AI models suggests China's threat to OpenAI and Anthropic may have been overblown
businessinsider.com · 3 November 2026Loading the full article…
This text was published by businessinsider.com and written by Alistair Barr. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Anthropic and OpenAI may be testing Fable 5.5 and GPT-6.2 Astra · 2 src
- Anthropic may have a new top-tier model in the works, report says · 2 src
- Anthropic reportedly prioritizes safety over image generation for Claude · 2 src
- LongCat debuts LongCat-2.5 with 1.6T parameters and visual token support · 1 src
- Deep Principle releases Hydrogen-1.0, a 753B open-weight science model · 1 src
Comments
via GitHub Discussions