{"rewrite":{"id":"r_92dd4dffb65c811ceebfa76e","clusterId":"c_1018e9d8ed7ab72c8e04446c","slug":"deepseek-releases-deepseek-v4-1-flash-1017-under-mit-license","model":"deepseek-v4-1-flash","headline":"DeepSeek Releases DeepSeek-V4.1-Flash-1017 Under MIT License","summary":"DeepSeek released the open model DeepSeek-V4.1-Flash-1017 on October 17, 2026, positioned as an updated, faster version of DeepSeek-V4.1-Flash. On the DeepSWE v1.1 coding agent benchmark it scores above GPT-6 Astra, GPT-6.1 Sol and Claude Fable 5 while keeping cost low, and it recorded the highest scores on multiple tests. Peak pricing is $0.3 per million input tokens and $1.2 per million output tokens, halved off-peak.","whyItMatters":"The coding benchmark chart and the MIT license put the price question on the table, since DeepSeek is selling usage cheaply while giving the weights away.","webCardHtml":"\u003cp\u003eThe model has 763 billion parameters, which DeepSeek says makes it impractical on a personal computer. Running it at usable speeds therefore depends on the official API or a third-party API, and Ollama and LM Studio have begun hosting it on servers in the US.\u003c/p\u003e\u003cp\u003ePeak-hour rates apply Monday through Friday, 1:00 to 4:00 and 6:00 to 10:00 UTC: $0.3 input, $0.006 cached input and $1.2 output per million tokens, with off-peak priced at half. The weights are downloadable from Hugging Face and ModelScope under the MIT License.\u003c/p\u003e","blueskyPost":"DeepSeek prices DeepSeek-V4.1-Flash-1017 at $0.30 per million input tokens at peak, halved off-peak. The benchmark win is notable; the pricing is what makes it usable at scale.","twitterPost":"DeepSeek's DeepSeek-V4.1-Flash-1017 undercuts on price as much as it wins on benchmarks, at $0.30 per million input tokens at peak.","threadsPost":"DeepSeek's MIT license on DeepSeek-V4.1-Flash-1017 is the part that travels. A benchmark score is a datapoint; an MIT license means anyone can run and redistribute the model, which is how it spreads beyond the leaderboard.","newsletterBlurb":"DeepSeek released DeepSeek-V4.1-Flash-1017 on October 17, 2026, an updated version of DeepSeek-V4.1-Flash that the company says runs faster and more accurately. On the DeepSWE v1.1 coding agent benchmark it scores above GPT-6 Astra, GPT-6.1 Sol and Claude Fable 5 while keeping cost low, and it took the highest score on multiple tests. The weights are free under the MIT License, though at 763 billion parameters practical use means the official or a third-party API.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261019-deepseek-v4-1-flash-1017/\",\"title\":\"Ultra-cheap Chinese AI model 'DeepSeek-V4.1-Flash-1017' arrives, beating GPT-6 Astra on coding benchmarks\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4913,"outputTokens":723,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1792379232,"createdAt":"2026-10-19T03:01:57.000Z","publishedAt":"2026-10-19T03:06:07.000Z","updatedAt":"2026-10-19T03:06:07.000Z"},"cluster":{"id":"c_1018e9d8ed7ab72c8e04446c","canonicalTitle":"コーディングベンチでGPT-6 Astraを超えつつ超安価な中国製AIモデル「DeepSeek-V4.1-Flash-1017」が登場","representativeArticleId":"a_cc19a1d331ca7f326ac524e6","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"DeepSeek-V4.1-Flash-1017\"],\"studios\":[\"DeepSeek\"],\"people\":[],\"type\":\"release\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-10-19T01:47:00.000Z","lastSeenAt":"2026-10-19T01:47:00.000Z","updatedAt":"2026-10-19T03:06:07.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20261019-deepseek-v4-1-flash-1017/","title":"コーディングベンチでGPT-6 Astraを超えつつ超安価な中国製AIモデル「DeepSeek-V4.1-Flash-1017」が登場"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["DeepSeek-V4.1-Flash-1017"],"studios":["DeepSeek"],"people":[],"type":"release","domain":"other","is_roundup":false},"keyFacts":null}