{"rewrite":{"id":"r_67275ab41215accf2a365eba","clusterId":"c_602cbc664b3e1734501edb53","slug":"deepseek-releases-v4-1-flash-as-an-open-model-with-lower-api-pricing","model":"deepseek-v4-flash","headline":"DeepSeek Releases V4.1-Flash As An Open Model With Lower API Pricing","summary":"DeepSeek released DeepSeek-V4.1-Flash on September 10, 2026. The open model is a mixture-of-experts system with 552 billion total parameters, 8 billion active parameters on input and 16 billion on output, and it supports image input. It beat Claude Opus 5 and GPT-5.6 Sol on multiple benchmark tests. Inference efficiency cut KV cache tokens to 1/3.9 of DeepSeek-V4-Flash, and off-peak API pricing is half the peak rate.","whyItMatters":"DeepSeek is giving away an MIT-licensed model that benchmarks above Claude Opus 5 and GPT-5.6 Sol on several tests and scores 40 on the Artificial Analysis Intelligence Index v4.3, which puts price pressure on Western labs that sell comparable performance per token.","webCardHtml":"\u003cp\u003eDeepSeek-V4.1-Flash is distributed on Hugging Face and ModelScope under the MIT License, with a technical report published alongside it. It uses pretraining and reinforcement learning methods that differ from DeepSeek-V4.\u003c/p\u003e\u003cp\u003eArtificial Analysis measured a score of 40 on its Intelligence Index v4.3, which was enough to pass GPT-5.6 Luna. On AutomationBench-AA, which tracks agent performance in clerical work such as finance and HR, the model recorded the top score ahead of GPT-6 Astra. Its output speed trails Gemini 3.8 Flash and Muse Spark 1.3, and it spends more output tokens per task than Claude Fable 5.1, so it is not the cheapest option per task among Chinese models. GLM-5.3 is slightly more cost-efficient.\u003c/p\u003e","blueskyPost":"DeepSeek's V4.1-Flash carries the same 552B parameters as larger competitors but activates only 8B on input, which is why its off-peak API rate runs half the peak rate.","twitterPost":"V4.1-Flash activates 8B of its 552B parameters on input. That slim active path is what lets DeepSeek halve off-peak API pricing.","threadsPost":"DeepSeek-V4.1-Flash is a 552B-parameter mixture-of-experts model that only wakes 8B of those parameters on input. The sparse activation is the cost story behind the half-rate off-peak API pricing, not the benchmark wins.","newsletterBlurb":"DeepSeek released DeepSeek-V4.1-Flash on September 10, an open mixture-of-experts model with 552 billion total parameters that it distributes under the MIT License and prices lower than its predecessor at the API level. It passed Claude Opus 5 and GPT-5.6 Sol on multiple benchmarks and scored 40 on the Artificial Analysis Intelligence Index v4.3. Third-party testing also put it on top of AutomationBench-AA, though output tokens per task run higher than Claude Fable 5.1.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260911-deepseek-v4-1-flash/\",\"title\":\"\\\"DeepSeek-V4.1-Flash\\\" arrives: an open model that outperforms Claude Opus 5 and GPT-5.6 Sol on multiple tests\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4834,"outputTokens":758,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1789096466,"createdAt":"2026-09-11T03:09:18.000Z","publishedAt":"2026-09-11T03:13:22.000Z","updatedAt":"2026-09-11T03:13:22.000Z"},"cluster":{"id":"c_602cbc664b3e1734501edb53","canonicalTitle":"「DeepSeek-V4.1-Flash」が登場、複数テストでClaude Opus 5やGPT-5.6 Solを超える性能のオープンモデル","representativeArticleId":"a_e7ddc01fa5a14e1b24bfd8fa","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"DeepSeek-V4.1-Flash\"],\"studios\":[\"DeepSeek\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-09-11T02:32:00.000Z","lastSeenAt":"2026-09-11T02:32:00.000Z","updatedAt":"2026-09-11T03:13:20.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260911-deepseek-v4-1-flash/","title":"「DeepSeek-V4.1-Flash」が登場、複数テストでClaude Opus 5やGPT-5.6 Solを超える性能のオープンモデル"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["DeepSeek-V4.1-Flash"],"studios":["DeepSeek"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":null}