模型多源确认

Nano Banana 2.1获Image Arena三项前六名

精选理由

Nano Banana 2.1在图像生成领域表现亮眼,Mistral Large 4在欧洲实验室中排名领先,Claude Haiku 5.5性价比突出。

Nano Banana 2.1在Image Arena三个模式中均进入前六:多图像编辑排名第4(1431分),文生图排名第5(1328分),图像编辑排名第6(1428分)。Mistral Large 4在Agent Arena排名第43位,较前版本提升11名。Claude Haiku 5.5(高配版)在WebDev代码竞技场排名第30位(1587分),价格与GPT-6 Luna相同但分数高6分。

图片来源 · lmarena.ai
原文 · lmarena.ai

This Week in the Arena: - Nano Banana 2.1 ranked in the top 6 across three Image Arena modes: #4 (1431 pts) in Multi-Image Edit, #5 (1328 pts) in Text-to-Image, and #6 (1428 pts) in Image Edit. - Mistral Large 4 landed at #43 overall in Agent Arena (-6.6%), ranking 11 spots above its previous variant, Mistral Medium 3.5 (-12.60%). The result also places Mistral AI among the top 15 labs, and it's the only European lab on the list! - Claude Haiku 5.5 (High) also landed in the Code Arena: WebDev at #30 (1587 pts). Just off the Pareto frontier, it matches GPT‑6 Luna’s price at $0.10/$0.50 per 1M input/output tokens, while scoring six pts higher. Agent score coming soon! - We also introduced Arena’s Alignment Index, and announced a $200M Series B at a $3.1B valuation. Find more info in thread. See the week’s leaderboards, research, and company updates in the full YouTube recap. youtube.com/watch?v=IHyMrt… 💬 9 🔄 0 ❤️ 45 👀 7125 📊 12 ⚡