OpenAI发布了名为“GPT-6.1 Terra”的新模型 1。Hacker News社区出现了关于该模型的讨论帖，链接指向OpenAI官方介绍页面，该帖子获得了17个点赞和3条评论 1。
根据相关介绍，GPT-6.1 Terra的主要功能包括匹配产品照片与目录条目、解读技术图纸、操作应用屏幕以及起草和修订文本 2。在技术参数上，该模型支持none、low、medium（默认）、high、xhigh和max六个推理努力级别 2。工具调用需要依赖Responses API 2。
在定价策略方面，缓存输入令牌的价格为非缓存输入令牌费率的5% 2。标准缓存写入价格分为两档：保留期为30分钟时，价格为每百万tokens 1.25美元；保留期为12小时时，价格为每百万tokens 2.50美元 2。对于长上下文请求，若输入令牌超过272K，输入和缓存费率将按2倍计算，输出费率则按1.5倍计算 2。此外，Batch和Flex模式的价格比Standard模式低50%，启用区域处理需加收10%费用，而Fast模式的价格是Standard模式的2倍 2。
OpenAI has introduced its latest artificial intelligence model, GPT-6.1 Terra 12. The model is designed to handle a variety of tasks, including matching product photos with catalog entries, interpreting technical drawings, operating application screens, and drafting or revising text 2. Tool calls for the new system require the use of the Responses API 2.
GPT-6.1 Terra supports a range of reasoning effort levels: none, low, medium (the default), high, xhigh, and max 2. The company has also outlined specific pricing strategies that vary based on context length, caching, and processing modes 2. Standard cache write prices are set at $1.25 per million tokens for a 30-minute retention period and $2.50 per million tokens for a 12-hour retention period, while cached input tokens cost 5% of the non-cached rate 2. For requests exceeding 272K input tokens, long-context adjustments apply: input and cache rates double (2x), and output rates increase by 1.5x 2.
Additional cost modifiers include a 10% premium for regional processing 2. Users can choose from different service tiers, where Batch and Flex modes are priced 50% lower than the Standard tier 2. Conversely, Fast mode incurs costs at twice the Standard price 2. The announcement has generated limited discussion on Hacker News, where one post linking to the official introduction received 17 points and three comments 1.
评论
还没有评论，欢迎留下第一条。