Loading...
Loading...
Gemini flash 3.5 昨晚发布,现已可用。 - 模型效果大幅超越 3.1 Pro,指标和 gpt 5.5 接近,比 gpt5.5 好的是 Agentic 和 多模态。 - 价格只要 gpt5.5 的三分之一,缓存价格只要六分之一。 - API 定价 $1.50 / $9.00 per 1M token(输入/输出),缓存 输入 $0.15。上下文窗口 1M token。 - 速度极快,是其他旗舰模型的4倍,非常适合 Agent 使用。 官方介绍地址: https://t.co/Sz2vB3O88p A new AI model reportedly detected pancreatic cancer up to three years earlier than human clinicians in retrospective tests, using medical imaging data to identify subtle early signals missed by doctors. The study analyzed historical scans a
Gemini 3.5 Flash introduces much faster multimodal inference and agentic capabilities at substantially lower cost than comparable flagship models, affecting architecture, deployment, and cost planning for AI services. Tech teams should reassess model selection, latency budgets, and token-cost economics when designing agents, multimodal apps, and inference pipelines.
Dossier last updated: 2026-05-20 01:59:39
试了下 Gemini 3.5 Flash,最大感受就是快
Simon Willison / Simon Willison's Weblog : Gemini 3.5 Flash costs $1.50 per 1M input tokens and $9 per 1M output tokens, 3x the price of Gemini 3 Flash Preview and 6x the price of Gemini 3.1 Flash-Lite — Today at Google I/O, Google released Gemini 3.5 Flash. This one skipped the -preview modifier and went straight to general availability …
Gemini flash 3.5 昨晚发布,现已可用。 - 模型效果大幅超越 3.1 Pro,指标和 gpt 5.5 接近,比 gpt5.5 好的是 Agentic 和 多模态。 - 价格只要 gpt5.5 的三分之一,缓存价格只要六分之一。 - API 定价 $1.50 / $9.00 per 1M token(输入/输出),缓存 输入 $0.15。上下文窗口 1M token。 - 速度极快,是其他旗舰模型的4倍,非常适合 Agent 使用。 官方介绍地址: https://t.co/Sz2vB3O88p
A new AI model reportedly detected pancreatic cancer up to three years earlier than human clinicians in retrospective tests, using medical imaging data to identify subtle early signals missed by doctors. The study analyzed historical scans and trained a machine-learning model to flag high-risk cases ahead of clinical diagnosis, potentially enabling earlier interventions in a cancer with poor prognosis. Key players include the research team developing the model and medical institutions that provided imaging datasets; specifics on model architecture, validation size, and peer review were not detailed. If validated prospectively and integrated into clinical workflows, the technology could shift screening practices and accelerate AI deployment in diagnostic radiology, but regulatory, privacy, and bias concerns remain.