摘要
新增了 GPT-6.1 Terra 的多个配置评测（High/Low/Max/Medium/Non-reasoning/Xhigh），以及新基准 Code2Video-AA 和文章 'GPT-6.1 Terra pushes the cost efficiency frontier'；删除了旧基准（AA-3D、FrontierCode）和 Muse 模型相关文章，显示评测重点转移。
LLM 生成LLM 摘要为机器生成，重要信息请回原文核对。
永久链接
https://daily.maas.click/item/obs_29624cd86d3733d5d1d6be35c8d6f94a7c5ef51d506dfe2bc2df9704c7c3c768/变化摘录
以下为已保存的变化摘录
证据级别：来源差异（source_diff）——这是抓取到的来源页面变化，其中定价页文本变化不代表已验证的调价。
新增 26 行
- Code2Video-AA
- Code2Video-AA: how well do AI models design videos in code?
- GPT-6.1 Terra (High)
- GPT-6.1 Terra (Low)
- GPT-6.1 Terra (Max)
- GPT-6.1 Terra (Medium)
- GPT-6.1 Terra (Non-reasoning)
- GPT-6.1 Terra (Xhigh)
- GPT-6.1 Terra pushes the cost efficiency frontier
- Muse Spark Open (Xhigh)See more
- New article published · 26 Oct
- New language model evaluation · 26 Oct
- Product and brand videos designed in code, Elo
- StepFunThinking MachinesXiaomiOpenAIAnthropicMistralMetaSpaceXAIKimiReflectionGoogleAlibabaZ AIMiniMaxNVIDIADeepSeek
- | Claude Fable 5.1 (Max, Default Fallback) | 4 (56) | 2 (60) | 2 (60) | 6 (50) | 3 (56) | 2 (63)
- | DeepSeek V4.1 Flash 1017 (Max) | 11 (48) | 11 (52) | 12 (46) | 10 (44) | 12 (45) | 14 (51)
- | GPT-6 Astra (Max) | 5 (55) | 5 (57) | 4 (59) | 3 (53) | 6 (55) | 6 (60)
- | GPT-6.1 Sol (Max) | 6 (54) | 6 (57) | 5 (58) | 4 (52) | 7 (54) | 7 (59)
- | Gemini 3.8 Flash (High) | 12 (47) | 12 (48) | 10 (52) | 15 (40) | 15 (43) | 12 (52)
- | Gemini 4 Argon (High) | 7 (54) | 7 (56) | 3 (60) | 9 (48) | 5 (55) | 4 (60)
- | Grok 4.7 (Xhigh) | 8 (52) | 10 (54) | 8 (54) | 8 (48) | 9 (47) | 9 (56)
- | Kimi K3.1 Pro | 3 (57) | 4 (59) | 7 (56) | 5 (52) | 4 (56) | 5 (60)
- | MiMo-V2.6-Pro | 13 (46) | 13 (48) | 15 (45) | 14 (41) | 10 (47) | 10 (53)
- | MiniMax-M3.1-Flash | 10 (49) | 9 (55) | 11 (47) | 11 (43) | 14 (43) | 15 (50)
- | Muse Wonder (Max) | 9 (51) | 8 (55) | 9 (52) | 7 (48) | 8 (54) | 8 (58)
- | Qwen3.8 Max (0902) | 14 (46) | 14 (47) | 14 (46) | 12 (41) | 11 (46) | 13 (52)
删除 6 行
- Announcing AA-3D: a benchmark for AI models building interactive 3D experiences
- Announcing FrontierCode on Artificial Analysis: can AI models write code maintainers would merge?
- Muse Spark Open (Xhigh)
- Muse Wonder Full (Max)See more
- Muse Wonder and Muse Wonder Full: a material jump in intelligence over previous Meta models
- Nemotron 3.5 Bolt 74B-A9B
修订历史（2 版）
- r1
- r2