跳到正文
Arena.ai· @arena · X·· 1 小时前AI 评分52
AI 导读

Arena 发布本周榜单更新。

正文 · 原文

This Week in the Arena: Open weights, faster image generation, and new #1s.

- @Kimi_Moonshot released the open weights for Kimi K3.1 Pro, which ranks #1 among open models in Agent Arena with +5.3% net improvement.

- Hy4 by @TencentHunyuan debuted as the #1 Chinese model in Agent Arena. With +5.4% net improvement at a $0.57 median cost per task, it also reshaped the Pareto frontier. Its open weights are coming soon.

- Inkling 2 Small by @thinkymachines arrived in Code Arena: WebDev with 1648 pts. Once its weights are released, it would rank #1 among US open models.

- Gemini Omni 1.2 Flash by @GoogleDeepMind debuted at #1 in both Text-to-Video (1541 pts) and Image-to-Video (1502 pts).

- GPT Image 2.6 Sunburst and Flare by @OpenAI arrived in the Arena at almost 2x the speed of their predecessor, GPT Image 2.5. Sunburst (Medium) landed at #2 in Text-to-Image (1409 pts) and #3 in Image Edit (1476 pts).

See more in the full update.

https://www.youtube.com/live/d1vJb7XZJPM

来源:Arena.ai · x.com