{
  "id": 13921676,
  "title": "THX-01 Just Processed 70 Million Tokens. Our Open-Source Decision Model Is Just Getting Started.",
  "url": "https://urgent.news/2026/10/12/thx-01-just-processed-70-million-tokens-our-open-source-decision",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-12T09:36:54.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/montydrief/thx-01-just-processed-70-million-tokens-our-open-source-decision-model-is-just-getting-started-2o8l"
  },
  "original_language": "en",
  "account": "THX-01, an open-source decision model with 322 million parameters, has processed over 70 million tokens since its launch. This model, designed to make AI decisions faster and more efficiently, specializes in providing structured answers to specific questions about documents or messages in a single forward pass. Unlike traditional AI applications that generate long responses, THX-01 focuses on making reliable decisions such as categorizing issues, identifying spam messages, or extracting numerical values. Evaluated against TypeSafe Jev, THX-01 achieved a higher accuracy of 98.4% compared to Jev's 97.4% while demonstrating significantly lower latency at approximately 10 milliseconds. Developed using Reinforcement Learning for Calibrated Decisions (RLCD), THX-01 provides calibrated probability estimates that reflect the model's actual uncertainty, making it particularly useful for AI agents and production applications requiring informed decision-making. Supported by 18 post-training languages, including Azerbaijani, English, Russian, and Turkish, THX-01 can also run locally on smartphones, making it accessible to a wider audience. The THX-01 API is free to use, with a generous free tier of 200 decisions per minute, ensuring that developers can experiment without incurring significant costs. The model, along with its source code and training resources, is openly available on Hugging Face, encouraging developers to contribute to and build upon this innovative technology. As HAL-X AI continues to advance the capabilities of THX-01 and the broader open-source AI ecosystem, the company remains committed to creating AI models that prioritize reliability, efficiency, and the ability to express uncertainty.",
  "summary": "70 million tokens processed. 322 million parameters. One forward pass. Free API. Fully open-source. A few days ago, we open-sourced THX-01 , our 322M-parameter non-autoregressive decision model designed to make AI decisions faster, more efficient, and accessible to everyone. Today, we've reached another milestone: over 70 million tokens processed through our THX-01 API since launch. And we're…",
  "key_points": [
    "THX-01 processed over 70 million tokens since launch",
    "Model achieves 98.4% accuracy, 10ms latency",
    "Open-source with 18 post-training languages"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "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."
}