Lakera Launches API to Protect Large Language Models from Malicious Attacks


DeepSeek officially announced that the API has been upgraded to the million-token context version, consistent with the App and web version. Previously, it was only 128k tokens. This update significantly improves the coherence of conversations and the richness of information. At the same time, the knowledge base has been updated to May 2025, allowing users to accurately access information up to that point even without an internet connection.
The efficiency of large language model inference has made a breakthrough. Tsinghua University and Moonshot AI jointly proposed a new architecture called "Prefill-as-a-Service," which splits the inference process into two stages: prefilling and decoding, and optimizes the allocation of computing resources, effectively solving hardware limitations and significantly improving model service performance.
Research shows that current mainstream AI models still have significant shortcomings in simulating clinical diagnostic reasoning and are not yet capable of independently handling medical tasks. This study tested 21 large language models, and the results were published in "JAMA Network Open".
Apple introduces two ML studies: SQUIRE enhances AI-generated UI control and fine-tuning with GPT-4o and slot query representation, while another improves image safety review to address current tech challenges.....
Alibaba's Qwen3.6-Plus tops OpenRouter's weekly global model call rankings, leading daily charts for four days. It achieved over 1 trillion tokens in daily calls, the first model on the platform to reach this milestone, highlighting the Qwen series' strong global competitiveness.....