{
  "id": 16592926,
  "title": "Euro open source specialist proposes LLM gateway to ease token bills",
  "url": "https://urgent.news/2026/10/26/euro-open-source-specialist-proposes-llm-gateway-to-ease-token-bills-16592926",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-26T10:15:00.000Z",
  "source": {
    "name": "The Register Science",
    "slug": "the-register-science",
    "url": "https://www.theregister.com/ai-and-ml/2026/10/26/euro-open-source-specialist-proposes-llm-gateway-to-ease-token-bills/5305170"
  },
  "original_language": "en",
  "account": "European open-source database provider Aiven has reportedly cut its AI token expenses by 80 percent by utilizing open-weight models and is developing a gateway to assist developers in seamlessly switching between various models as needed. Aiven, headquartered in Finland, initially focused on providing open-source databases as a service, but has since expanded into machine learning. According to Ville Lehto, Aiven's VP of strategy, creating an LLM infrastructure compatible with multiple clouds is not a straightforward task. However, Aiven aimed to control its own destiny and reduce costs when employing AI. Following a discussion on token-maxxing, developers have gravitated towards cost-saving strategies, incorporating a blend of models, such as budget-friendly Chinese options versus open-weight models that are less expensive. Open-weight models have openly published learned parameters, such as weights and biases, and can operate locally or be hosted at a lower cost compared to high-end models from companies like OpenAI and Anthropic. Examples include Meta's Llama Family, Alibaba Cloud's Qwen Family, China's DeepSeek, Mistral AI's Medium 3.5, and Devstar and Google's Gemma. Lehto mentioned that Aiven opted for a more efficient token utilization approach by constructing an internal gateway to simplify the handling of open-weight models for its developers. The company focused on heavy users, achieving an approximate 80 percent reduction in expenses. While frontier models are still employed for planning purposes, the majority of tasks are completed using open-weight models. Aiven intends to provide this gateway as a service to customers to assist them in managing token costs by facilitating the effortless substitution of open-weight models within their workflows. Aiven's clientele comprises German e-commerce firm Conrad, UK payment processing company Dojo, cybersecurity firm Sophos, and communications company Avaya.",
  "summary": "Aiven hopes to do for LLMs what it does for databases",
  "key_points": [],
  "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."
}