{"rewrite":{"id":"r_c940e4388229bc27e2584723","clusterId":"c_9bb66a164bec31188b6da8eb","slug":"fujitsu-develops-photon-a-transformer-alternative-that-boosts-ai-efficiency-475-times","model":"deepseek-v4-flash","headline":"Fujitsu Develops PHOTON, a Transformer Alternative That Boosts AI Efficiency 475 Times","summary":"Fujitsu announced PHOTON, a new architecture for large language models that improves throughput per GPU resource by up to 475 times compared to Transformer. PHOTON processes text in meaningful chunks rather than token units and uses multi-query integration for stable performance. The company will present details at ACL 2026 in July.","whyItMatters":"PHOTON directly addresses Transformer's memory bottleneck for long documents and multi-user queries, offering a path to much cheaper inference for multi-agent systems and large-scale LLM deployment.","webCardHtml":"\u003cp\u003eFujitsu\u0026#39;s new architecture, PHOTON, treats text as hierarchical chunks instead of individual tokens, cutting the computational load that makes Transformer slow on long inputs or simultaneous queries. In tests with a 1.2 billion parameter model, PHOTON achieved up to 475 times the multi-query throughput per GPU resource compared to standard Transformer, with only slight performance degradation. The architecture also uses a multi-query integration technique that combines multiple candidate outputs via majority voting or best-candidate selection, producing stable results from a single inference pass. Fujitsu plans to present the full details at ACL 2026, the top natural language processing conference, running July 2-7 in San Diego.\u003c/p\u003e","blueskyPost":"Fujitsu's PHOTON architecture claims up to 475x throughput per GPU vs Transformer by processing text in chunks instead of tokens. Details at ACL 2026 next week.","twitterPost":"Fujitsu's PHOTON architecture claims up to 475x throughput per GPU vs Transformer by processing text in chunks instead of tokens. Details at ACL 2026 next week.","threadsPost":null,"newsletterBlurb":"Fujitsu announced PHOTON, a new LLM architecture that processes text in hierarchical chunks rather than token units, achieving up to 475 times the throughput per GPU resource compared to Transformer. The design also integrates multiple query outputs for stable single-inference performance. Fujitsu will present the work at ACL 2026 in July.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260625-fujitsu-photon/\",\"title\":\"Fujitsu Develops 'PHOTON,' a Transformer Alternative Architecture That Improves AI Efficiency by 475 Times\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4169,"outputTokens":538,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1782593788,"createdAt":"2026-06-27T20:52:58.000Z","publishedAt":"2026-06-27T20:55:24.000Z","updatedAt":"2026-06-27T20:55:24.000Z"},"cluster":{"id":"c_9bb66a164bec31188b6da8eb","canonicalTitle":"富士通がAI効率を475倍にするTransformer代替アーキテクチャ「PHOTON」を開発","representativeArticleId":"a_a968be7262e7a50ee6cf6bf5","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-06-25T02:08:00.000Z","lastSeenAt":"2026-06-25T02:08:00.000Z","updatedAt":"2026-06-27T20:55:24.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260625-fujitsu-photon/","title":"富士通がAI効率を475倍にするTransformer代替アーキテクチャ「PHOTON」を開発"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":["Fujitsu's PHOTON architecture improves multi-query throughput per GPU resource by up to 475 times compared to standard Transformer in tests with a 1.2 billion parameter model.","PHOTON processes text as hierarchical chunks rather than individual tokens, reducing computational load for long documents and simultaneous queries.","The architecture uses multi-query integration, combining multiple candidate outputs via majority voting or best-candidate selection to produce stable results from a single inference pass.","Fujitsu will present full details of PHOTON at ACL 2026, the top natural language processing conference, running July 2-7 in San Diego."]}