HappyHorse-1.0 Tops AI Video Arena at 1383 Elo
Video generation model HappyHorse-1.0 ranks No.1 on Artificial Analysis’ AI Video Arena. It achieved an Elo score of 1383. Developer and technical details are undisclosed.
Search
Tag: #leaderboard25 results
Qwen 3.6 Plus Tops Global LLM Calls Chart
Alibaba's Qwen 3.6 Plus has secured the top position in the global large model invocation weekly leaderboard. A more powerful flagship model, Qwen 3.6 Max, is slated for upcoming release.
Qwen 3.6 Plus Hits 1.4T Daily Tokens, Tops Global List
Alibaba's Qwen 3.6 Plus has shattered records with over 1.4 trillion daily token invocations. It now leads the global model usage leaderboard. This highlights its massive adoption and performance at scale.
Feb 2026 SWE-rebench: Claude Tops at 65.3%
The February 2026 SWE-rebench leaderboard update features 57 fresh GitHub PR tasks, with Claude Opus 4.6 leading at 65.3% resolved rate. Top models like GPT-5.2-medium (64.4%), GLM-5 and GPT-5.4-medium (both 62.8%) form a tight race. Open-weight models such as Qwen3.5-397B (59.9%) are rapidly closing the gap.
Ungameable LLM Leaderboard Funded by Ranked Companies
Arena, formerly LM Arena, has become the de facto public leaderboard for frontier LLMs, influencing funding, launches, and PR cycles. Funded by the companies it ranks, the startup evolved from a UC Berkeley PhD research project in just seven months.
PhD Students Judge AI via Arena Leaderboard
Arena, formerly LM Arena, has become the de facto public leaderboard for frontier LLMs, founded by UC Berkeley PhD students. In just seven months, it evolved from a research project into a startup influencing AI funding, launches, and PR cycles amid fierce model competition.
📰
ColQwen3.5-v3 Tops ViDoRe Leaderboard
ColQwen3.5-4.5B-v3 leads MTEB ViDoRe leaderboard at 75.67 mean, using half params and 13x fewer embedding dims than prior #1. Full eval trail public; supported by colpali-engine and vLLM on ROCm/CUDA. Apache 2.0, with 9B variant training.
🔬
IDP Leaderboard Benchmarks 16 VLMs
New open IDP Leaderboard evaluates 16 VLMs on 9,000+ documents across three benchmarks: OlmOCR, OmniDoc, and IDP Core. Gemini 3.1 Pro leads narrowly; cheaper variants match flagships except on reasoning tasks. Features a Results Explorer for predictions vs. ground truth.
Step3.5 Flash Tops OpenClaw for 3 Days
階躍星辰的Step3.5 Flash模型根據OpenRouter數據，已連續三天在OpenClaw上保持全球調用量第一。2026年3月以來，Kimi K2.5、Step3.5 Flash與MiniMax M2.5的調用量分居全球前三。
Independent developer yuxinlu1 climbs Hugging Face leaderboard
An independent developer known as yuxinlu1 has achieved a top ranking on the Hugging Face model leaderboard. This feat highlights the growing competitiveness of individual contributors against major tech corporations.