{
  "id": 15684281,
  "title": "Inception launches Mercury 3, a diffusion-based LLM it claims is \"comparable to Luna and Haiku models\" but 4x to 8x faster, generating ~1,000 tokens per second (Jason Hiner/The Deep View)",
  "url": "https://urgent.news/2026/10/20/inception-launches-mercury-3-a-diffusion-based-llm-it-claims-is",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-20T01:45:01.000Z",
  "source": {
    "name": "Techmeme",
    "slug": "techmeme",
    "url": "https://www.thedeepview.com/articles/why-inception-s-mercury-3-upends-ai-economics"
  },
  "original_language": "en",
  "account": null,
  "summary": "Inception has launched Mercury 3, a diffusion-based large language model. According to Hacker News, the model is the most capable diffusion-based LLM released to date, outperforming GPT-6 Luna and Claude Haiku 5.5 across various benchmarks. Techmeme reports the model is comparable to Luna and Haiku models but runs 4x to 8x faster.\n\nHacker News states that Mercury 3 generates 1,068 tokens per second on NVIDIA GPUs and features a 1M token context window. The model is more than 40x larger than Mercury 1, with intelligence rising 3x across standard benchmarks compared to the first version. Early-access partners are utilizing the model for tasks such as web navigation, data engineering, real-time tutoring, and voice and chat support.\n\nThe standard pricing for Mercury 3 is $0.20 per million input tokens and $0.70 per million output tokens, though it is currently 50% off at launch. Hacker News notes that cached tokens are available at 90% off the input price.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Hacker News",
        "title": "Inception \"Mercury 3\", 1000 tok/s diffusion LLM",
        "url": "https://urgent.news/2026/10/19/inception-mercury-3-1000-tok-s-diffusion-llm",
        "published": "2026-10-19T17:27:11.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}