{"rewrite":{"id":"r_77a1c9a96f9644d5a9e39e55","clusterId":"c_fd780c3a3f114f37a0f26cd4","slug":"simon-willison-s-pelican-test-marks-november-2025-as-ai-inflection-point","model":"deepseek-v4-flash","headline":"Simon Willison's Pelican Test Marks November 2025 as AI Inflection Point","summary":"Developer Simon Willison presented his 'pelican riding a bicycle' benchmark results at PyCon US 2026, tracking model improvements from November 2025 to May 2026. He identified November 2025 as an inflection point, with models like Gemini 3, GPT-5.1, and Claude Sonnet 4.5 showing increasingly natural pelican-bicycle relationships, while open-weight models like Qwen3.6-35B-A3B matched top-tier performance.","whyItMatters":"The informal benchmark provides a concrete, visual measure of how quickly image generation quality improved across both proprietary and open-weight models in just six months.","webCardHtml":"\u003cp\u003eSimon Willison has been running the same test since June 2025: ask a large language model to draw a pelican riding a bicycle. At PyCon US 2026, he presented the results from November 2025 through May 2026, calling the start of that period an inflection point. Early outputs from Claude Sonnet 4.5 and GPT-5.1 showed distorted pelican-bicycle relationships. By February 2026, Gemini 3.1 Pro produced a natural image, though it added an unprompted fish in the basket. Open-weight models also improved. Qwen3.6-35B-A3B, a 20.9 GB model that runs on a laptop, generated a more natural image than Anthropic's top-tier Claude Opus 4.7. Willison noted that GLM-5.1, a 1.51 TB open-weight model, handled the still image well but produced distorted motion when asked to animate it.\u003c/p\u003e","blueskyPost":"Willison's pelican-on-bicycle benchmark shows November 2025 was an inflection point. By May 2026, a laptop-sized open-weight model beat a top-tier Anthropic model on the same test.","twitterPost":"Willison's pelican-on-bicycle benchmark: November 2025 was the inflection point. By May 2026, a 20.9 GB open-weight model running on a laptop beat Anthropic's top Claude Opus 4.7 on the same test.","threadsPost":null,"newsletterBlurb":"Simon Willison updated his informal AI benchmark at PyCon US 2026, tracking how models draw a pelican riding a bicycle. He identified November 2025 as an inflection point, with Gemini 3.1 Pro producing natural images by February 2026. Open-weight models like Qwen3.6-35B-A3B matched or exceeded top-tier proprietary models by May 2026.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260521-llms-pelicans-on-bicycles-2026/\",\"title\":\"This is what happens when you have Gemini 3.1 Pro and Qwen3.6-35B-A3B draw a pelican riding a bicycle for the AI benchmark\"}]","lintFlagsJson":"[]","lintHits":0,"costUsd":0,"inputTokens":3950,"outputTokens":682,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1779826594,"createdAt":"2026-05-20T23:00:48.000Z","publishedAt":"2026-05-20T23:15:01.000Z","updatedAt":"2026-05-20T23:15:01.000Z"},"cluster":{"id":"c_fd780c3a3f114f37a0f26cd4","canonicalTitle":"AIベンチマーク「自転車に乗ったペリカンを描く」をGemini 3.1 ProやQwen3.6-35B-A3Bにやってもらうとこうなる","representativeArticleId":"a_31a17c781f93cdb4ecfe30d4","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"studios\":[],\"people\":[\"サイモン・ウィリソン\"],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-05-20T21:00:00.000Z","lastSeenAt":"2026-05-20T21:00:00.000Z","updatedAt":"2026-05-20T23:15:01.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260521\u0026#45;llms\u0026#45;pelicans\u0026#45;on\u0026#45;bicycles\u0026#45;2026/","title":"AIベンチマーク「自転車に乗ったペリカンを描く」をGemini 3.1 ProやQwen3.6-35B-A3Bにやってもらうとこうなる"}],"entities":{"anime_titles":[],"manga_titles":[],"studios":[],"people":["サイモン・ウィリソン"],"type":"news","domain":"other","is_roundup":false},"keyFacts":["Simon Willison presented his 'pelican riding a bicycle' benchmark results at PyCon US 2026, tracking model improvements from November 2025 to May 2026.","Willison identified November 2025 as an inflection point, with models like Gemini 3, GPT-5.1, and Claude Sonnet 4.5 showing increasingly natural pelican-bicycle relationships.","By February 2026, Gemini 3.1 Pro produced a natural image of a pelican riding a bicycle, though it added an unprompted fish in the basket.","Open-weight model Qwen3.6-35B-A3B, a 20.9 GB model that runs on a laptop, generated a more natural image than Anthropic's top-tier Claude Opus 4.7.","GLM-5.1, a 1.51 TB open-weight model, handled the still image well but produced distorted motion when asked to animate it."]}