Loading / 加载中
A brief snippet from a Chinese report states that a third-party evaluation has found a model to surpass OpenAI's strongest flagship, GPT-5.6 Sol. No specific model name, benchmark, or institution is provided in the snippet. The claim remains unverified and lacks detail.
The article consists solely of the Chinese phrase '真·有求必应·阿拉丁' ('True, responsive, Aladdin'), a playful remark that appears to be in response to a call for opinions on GLM-5.3. No concrete information about any model, release, features, or feedback details is provided. The content does not support the title's implication of a global opinion solicitation or vision-related demands.
DeepSeek has announced the official version of V4. The new version will feature upgraded capabilities compared to previous releases. Specific performance improvements or new features were not detailed in the announcement. The release marks the transition from earlier V4 previews to a finalized product.
DeepSeek founder Liang Wenfeng has decided to pursue external financing. The move was directly prompted by competitive pressure from Anthropic's Claude Mythos model. The company signaled that without fresh capital it cannot sustain the fight. No details on the financing round or amount were disclosed.
The author states that they personally created a Xianxia-themed short drama. No further details about the production tools, techniques, or any AI involvement are provided.
Baidu has released a new open-source OCR system that claims to process an entire book in a single pass. The tool is reportedly developed by a researcher who previously worked at DeepSeek. No further technical details, model name, or benchmark results were provided in the source.