{"version":1,"type":"story","url":"https://digestai.news/story/fortytwo-launches-engine-api-cutting-tokens-by-24-7-on-coding-tasks","json":"https://digestai.news/story/fortytwo-launches-engine-api-cutting-tokens-by-24-7-on-coding-tasks.json","markdown":"https://digestai.news/story/fortytwo-launches-engine-api-cutting-tokens-by-24-7-on-coding-tasks.md","slug":"fortytwo-launches-engine-api-cutting-tokens-by-24-7-on-coding-tasks","headline":"Fortytwo introduces swarm inference to cut AI token use by 24.7%","summary":"Fortytwo, founded by Ivan Nikitin and Vlad Larin, has developed swarm inference technology to improve AI model accuracy and efficiency. The company’s **Fortytwo Engine** uses a **Swarm Decoder** to let multiple models collaborate in real time, evaluating and steering reasoning as it unfolds. This reduces unnecessary token generation and improves task completion rates, particularly on long, complex tasks like software engineering.\n\nNikitin, who previously co-founded **Temporal** and worked on conversational AI at **Sensorium**, explains that traditional single-model approaches struggle with **model drift**—where context becomes cluttered with irrelevant or erroneous outputs. Fortytwo’s swarm architecture catches errors early, preventing wasted compute. Testing with **Qwen 3.8 27B** on **450 DeepSWE tasks** showed a **24.7% reduction in token usage** while improving success rates from **42.2% to 47.8%**. The Engine is **OpenAI API-compatible**, making it easy for developers to integrate without rewriting code.","keyPoints":["Fortytwo Engine uses swarm inference to reduce token use by 24.7% while improving task accuracy on long AI runs","Real-time peer evaluation prevents model drift, cutting wasted compute on tasks like software engineering","Engine is OpenAI API-compatible, allowing seamless integration for developers without code changes"],"whyItMatters":"Swarm inference could redefine AI efficiency by shifting focus from model size to reasoning optimization, cutting costs for developers running long, complex tasks.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":["Fortytwo","Temporal","Sensorium"],"models":["Qwen 3.8 27B"],"people":["Ivan Nikitin","Vlad Larin"]},"firstPublishedAt":"2026-10-15T14:00:05Z","updatedAt":"2026-10-15T14:03:48Z","sourceCount":2,"hasPrimarySource":false,"sources":[{"outlet":"Unite.AI","title":"Ivan Nikitin, CEO and Co-Founder of Fortytwo","url":"https://unite.ai/ivan-nikitin-ceo-and-co-founder-of-fortytwo-interview-series","publishedAt":"2026-10-15T14:03:48Z","type":"press","primary":false,"lead":true},{"outlet":"Unite.AI","title":"Fortytwo Launches Engine API, Reporting 24.7% Fewer Tokens on Coding Tasks","url":"https://unite.ai/fortytwo-launches-engine-api-reporting-24-7-fewer-tokens-on-coding-tasks","publishedAt":"2026-10-15T14:00:05Z","type":"press","primary":false,"lead":false}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Fortytwo introduces swarm inference to cut AI token use by 24.7%\", 15 October 2026, https://digestai.news/story/fortytwo-launches-engine-api-cutting-tokens-by-24-7-on-coding-tasks","publisher":"Digest AI","title":"Fortytwo introduces swarm inference to cut AI token use by 24.7%","datePublished":"2026-10-15T14:00:05Z","url":"https://digestai.news/story/fortytwo-launches-engine-api-cutting-tokens-by-24-7-on-coding-tasks"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}