Inception 发布扩散语言模型 Mercury 3，每秒输出超 1000 tokens
Inception 的 Mercury 3 用扩散方式生成文本，每秒 1000+ tokens，号称能打 GPT-6 Luna 和 Claude Haiku 5.5，做语音客服这类场景很合适。
Inception 发布扩散语言模型 Mercury 3，规模是 Mercury 1 的 40 倍以上，在标准基准上智能水平提升 3 倍。该模型输出速度仍保持在每秒 1000 tokens 以上，配备 100 万 token 上下文窗口。Inception 称其在多项基准上对标或超越 OpenAI 的 GPT-6 Luna 和 Anthropic 的 Claude Haiku 5.5。与传统逐词生成的 ChatGPT 和 Claude 不同，Mercury 3 采用一次生成多词再逐步精炼的扩散机制，适合语音智能体、实时辅导和客服等对延迟敏感的场景。模型已在 Inception API、Baseten 和 OpenRouter 上线。
Inception just launched Mercury 3, which it calls the strongest diffusion language model yet. It’s 40x bigger than Mercury 1 but still puts out 1,000+ tokens a second. The comparison is to the fast, cheap tier, with Inception saying it matches or beats OpenAI’s GPT-6 Luna and Anthropic’s Claude Haiku 5.5. Where ChatGPT and Claude write one word at a time, diffusion models draft many words at once and refine them, making them a strong fit for use cases like voice agents, live tutoring, and customer support, where any lag shows. Inception @_inception_ai Mercury 3 is live: the most capable diffusion LLM released to date. It outperforms GPT-6 Luna and Claude Haiku 5.5 across a range of benchmarks while running at over 1,000 tokens per second. More than 40x the size of Mercury 1, 3x more intelligent across standard benchmarks, and just as fast, with a 1M-token context window. Available today on the Inception API, @baseten , and @OpenRouter . 🔗 View Quoted Tweet 💬 4 🔄 0 ❤️ 4 👀 1852 📊 3 ⚡
