Tag: #leaderboard25 results
The world's first large model specialized in medical video understanding has been open-sourced. It launches alongside over 6,000 high-quality annotated test sets and a public leaderboard. Developers are encouraged to test and contribute.
Xiaomi CEO Lei Jun announced MiMo large model surpassed 1 trillion token calls. MiMo-V2-Pro ranked top 5 globally on Text Arena for logic, instruction following, and dialogue. It dominates OpenRouter daily, weekly, and trend leaderboards.
ClawBench released its latest large model leaderboard, with Zhipu, ByteDance, and Xiaomi securing 4 spots in the global top 10. Zhipu GLM-5-Turbo leads with 93.9 score, ByteDance Doubao-Seed-2.0-lite is #2 with lowest cost, and Xiaomi MiMo-V2-Omni ranks #9 as the fastest.
Alibaba's Qwen3.5-Max-Preview has debuted on LMArena, topping the rankings among global AI models. It outperforms both international leaders and domestic competitors, signaling strong performance gains.
Microsoft's MAI-Image-2 ranks #3 on Arena.ai's leaderboard, trailing only Google and OpenAI models. It begins rolling out today across Copilot and Bing Image Creator. This marks a shift from relying on OpenAI's image generation tech.
Kunlun Tech's SkyReels-V4 model has secured the top position globally in the 'Text to Video (With Audio)' category on the Artificial Analysis leaderboard. It outperforms major competitors including Kling 3.0, Google Veo 3.1, and OpenAI Sora 2.
Xiaomi has claimed the No.1 mysterious model Hunter Alpha on a leaderboard. The model features trillion parameters and a million-token context window. Even prominent figure 'Father of Lobster' is inquiring about it.
The crowdsourced AI leaderboard, originally a UC Berkeley research project, has scaled to $100 million in annualized revenue. This milestone was achieved just eight months after the launch of its first commercial product.
Papers With Code has updated its platform to include performance evaluations for closed-source models like GPT-5.5 and Mythos 5. Users can now toggle between open-source and closed-source leaderboards to better compare state-of-the-art AI performance.
Hugging Face introduces QIMMA, a new leaderboard focused on evaluating Arabic Large Language Models with an emphasis on quality metrics. It aims to highlight top-performing Arabic LLMs in a standardized manner. This tool supports advancements in Arabic NLP research and development.