{"slug": "celeris-1-magnus-fast-hybrid-diffusion-model-for-agentic-work", "title": "Celeris-1 Magnus: Fast hybrid diffusion model for agentic work", "summary": "Celeris AI released Celeris-1 Magnus, a hybrid diffusion model for agentic work, scoring 41.2% on τ³-bench banking, ahead of gpt-5.6-sol, with a 13.4-point gain when thinking is enabled for 6 seconds and a 55-second median completion time per task. The model is available via an OpenAI-compatible API at inference.celeris.ai, allowing existing agent code to run with a one-word change.", "body_md": "Magnus is built for agents that need to think, use tools, and get things done, without waiting around.\n\nAll 97 tasks of τ³-bench banking, head-to-head against gpt-5.6-sol, gpt-5.6-luna and gemini-3.7-flash, reasoning effort as published.\n\n41.2% on τ³-bench banking, ahead of gpt-5.6-sol, with the field's best solve rate.\n\nOne flag between fast and thorough. 13.4 extra points when thinking is on, for 6 seconds.\n\n55 seconds median per completed task, quicker than every model within ten points of it.\n\nCeleris-1 is built for answers; Magnus is built for tasks. The API is the same, so switching is a one-word change.\n\nUltra-fast generation for chat, search and real-time interfaces.\n\nAgentic work: tool loops, long tasks, structured actions.\n\nThe API is OpenAI-compatible. Point your SDK at inference.celeris.ai and keep the agent code you already wrote.\n\nSign up at console.celeris.ai. No waitlist, no sales call.\n\nKeys are live immediately and billed per token.\n\nSet the base URL and model id. Your existing agent code runs as is.\n\n``` python\nfrom openai import OpenAI\n\nclient = OpenAI(\n    base_url=\"https://inference.celeris.ai/celeris-1-magnus/v1\",\n    api_key=\"ck_…\",\n)\n\nresp = client.chat.completions.create(\n    model=\"celeris-1-magnus\",\n    messages=[{\"role\": \"user\", \"content\": \"Move my Friday payment to Monday\"}],\n    tools=my_tools,\n    # the reasoning dial: on when the task earns it\n    extra_body={\"chat_template_kwargs\": {\"enable_thinking\": True}},\n)\n```\n\n", "url": "https://wpnews.pro/news/celeris-1-magnus-fast-hybrid-diffusion-model-for-agentic-work", "canonical_source": "https://celeris.ai/celeris-1-magnus", "published_at": "2026-09-01 05:20:08+00:00", "updated_at": "2026-09-01 05:52:20.315079+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products", "ai-agents"], "entities": ["Celeris AI", "Celeris-1 Magnus", "gpt-5.6-sol", "gpt-5.6-luna", "gemini-3.7-flash", "OpenAI"], "alternates": {"html": "https://wpnews.pro/news/celeris-1-magnus-fast-hybrid-diffusion-model-for-agentic-work", "markdown": "https://wpnews.pro/news/celeris-1-magnus-fast-hybrid-diffusion-model-for-agentic-work.md", "text": "https://wpnews.pro/news/celeris-1-magnus-fast-hybrid-diffusion-model-for-agentic-work.txt", "jsonld": "https://wpnews.pro/news/celeris-1-magnus-fast-hybrid-diffusion-model-for-agentic-work.jsonld"}}