GoogleDeepMind发布Gemini 3.8 Flash (High)，在多个基准测试中超越前代版本和Claude Opus 5。
Gemini 3.8 Flash (High)在Agent Arena排名第14，较3.7版本提升5.94个百分点。在Text Arena中位列第7，超越Claude Opus 5 (High)的1492分。该模型在写作、多轮对话和长查询等类别中排名显著提升。
Gemini 3.8 Flash (High) by @GoogleDeepMind is here! It just debuted across Agent Arena, Text Arena, ...
Gemini 3.8 Flash (High) by @GoogleDeepMind is here! It just debuted across Agent Arena, Text Arena, and Code Arena: WebDev. In Agent Arena, it landed #14 with +5.94% net improvement. This ranks just above DeepSeek-V4-Pro at #15 (+5.91%) and is a significant jump from Gemini 3.7 Flash (High) at #32 (+0.84%). Its strongest signals are: +14.78% in Praise vs. Complaint (implicit sentiment in users reactions) +11.12% in Confirmed Success (explicit “yes, that worked” from users) In Text Arena, Gemini 3.8 Flash (High) is #7 with 1494 pts, ahead of Claude Opus 5 (High) at #8 (1492 pts) and Gemini 3.7 Flash (High) at #10 (1491 pts)! This release improved over Gemini 3.7 Flash (High) by category as well: - Writing, Literature & Language: #7 → #3 - Multi-Turn: #9 → #4 - Longer Query: #13 → #5 - Hard Prompts (English): #23 → #6 - Business, Management & Financial Ops: #26 → #7 - Coding: #22 → #7 - Hard Prompts: #13 → #7 - Instruction Following: #11 → #9 - Software & IT Services: #18 → #12 In the Code Arena: WebDev, Gemini 3.8 Flash (High) is #18 with 1567 points, enough to keep Google DeepMind at #8 when labs are ranked by their best-performing model. Congrats to the @GoogleDeepMind team on the release! Google DeepMind @GoogleDeepMind Two new Gemini models are here to help scale your AI agents and secure code: 🔘 3.8 Flash: our most intelligent model yet with significant gains from 3.7 Flash across software engineering, agentic tasks, and multi-step reasoning. 🔘 3.8 Flash Cyber: our most capable cybersecurity model with frontier-level vulnerability detection and automated patching. 🔗 View Quoted Tweet 💬 10 🔄 6 ❤️ 124 👀 10891 📊 18 ⚡