待翻譯：DeepSeek v4 Price Increase

AI 服務暫時不可用，以下為來源正文，待恢復後補全翻譯。

DeepSeek (@deepseek_ai): "API pricing update 💰 With the V4 lineup release, we’re updating our API pricing and introducing peak and off-peak rates. Off-peak rates are 50% lower than peak, enabling more flexible workload scheduling. 📉 New pricing takes effect at 16:00 UTC, Aug 16, 2026 🕒" | XCancel @deepseek_ai 1h We’re launching DeepSeek-V4-Pro today! 🚀 🔷 Major Agent upgrades with strong production gains! 🔷 Flexible reasoning effort for V4-Pro & V4-Flash: low for simple tasks, high for daily Agent workflows, max for complex tasks. 🔷 Native OpenAI Responses API support, optimized for Codex with one-click setup. V4 Pro is now available on app/web. Try it via “Expert Mode”. V4 Pro is also available via API. Model names remain unchanged—please refer to the API docs for setup details. 339 832 6,687 358,249 @deepseek_ai 1h API pricing update 💰 With the V4 lineup release, we’re updating our API pricing and introducing peak and off-peak rates. Off-peak rates are 50% lower than peak, enabling more flexible workload scheduling. 📉 New pricing takes effect at 16:00 UTC, Aug 16, 2026 🕒 Aug 13, 2026 · 11:31 AM UTC 104 86 834 98,264 Sort replies: Relevant Recent Liked 52m Replying to @deepseek_ai GPT 5.6 Luna (vision + same intelligence + much lower price) >>> deepseek v4 flash (no vision + same intelligence + much higher price) 387 @ssbengale 13m Replying to @deepseek_ai Pretty reasonable pricing. I hope the other providers maintain the prices for Flash atleast. That model at the old prices would be enough for me. 1 421 21m Replying to @deepseek_ai Still no llm vision before the inflation chokes the throat. 323 @MarcelHolter 26m Replying to @deepseek_ai Isn’t V4 Flash way more expensive in off peak hours than it was before? Wasn’t it 0.14/0.0028/0.28? 898 1h Replying to @deepseek_ai Please atleast leave flash as it used to be. Deepseek should not be Fable 5 competitor, rather a super cheap, good quality overall ai. 7 1,346 @kenrin_0 1h Replying to @deepseek_ai Still insane cheap though if the claims hold up though. 2,031 42m Replying to @deepseek_ai This is great and all, but we need a deepseek mini version. A cheaper model for simpler tasks, like translation. Machine translation is not great and is expensive on Google / Microsoft / etc. AI translation is much better, but also expensive. 152 @JinseokKim60030 1h Replying to @deepseek_ai The 30x cache-hit vs miss gap on V4-Pro is the whole ballgame for agent loops. Most cost dashboards I've seen still lump input into one bucket, so a long CC/Codex session looks 3-4x more expensive than it was. Peak/off-peak on top means the same run prices differently by hour. 1 734 41m Replying to @deepseek_ai After this pricing update i’ll stick to GPT Luna . Thanks 91 @foodtruckbench 1h Replying to @deepseek_ai Damn, this is going to be expensive. Up to 17 times more expensive than the Luna in our benchmark. 1 3 1,288 9m Replying to @deepseek_ai Bold of you to assume there’ll still be a peak hour at these prices 😭 197 @VincentWong 1h Replying to @deepseek_ai @grok can you compare the deepseek v4 flash with other competitors like gpt-5.6 luna, grok 4.6, claude sonnet 5, muse spark 1.2 2 1,747 @btsouth 1h Replying to @deepseek_ai It was fun while it lasted. There isn't really a reason to use DeepSeek anymore. 5 72 4,518 @JeremySmit48006 29m Replying to @deepseek_ai There’s no reason to use pro at all after the increase. Looks like Flash for me till I start running locally 24/7 1 436 1h Replying to @deepseek_ai @grok with the new pricing is deepseeks official API the cheapest option any longer? Or is there now cheaper options from third-party providers for deepseek v4 flash? 1 1,142 @justusvv 1h Replying to @deepseek_ai @grok how do the new prices compare to the current prices? 2 4 6,032 31m Replying to @deepseek_ai Ds v4 pro compared with the previous price, there will be many fewer people using it. At present, this model has little improvement 807 1h Replying to @deepseek_ai Why the fuck do you subsidizes US and EU and increase price for everyone near ASIAN time zone!!! It should be the opposite because US and EU has more money 9 122 7,581 1h Replying to @deepseek_ai unfortunately these new prices made v4 completely unappealing to me. 3 44 3,280 1h Replying to @deepseek_ai @CommandCodeAI - How does this affect your GOAT plan? 3 2,048 1h Replying to @deepseek_ai It was a pleasure... Bye DeepSeek. I am going to burn my last dollars in the next days! 10 778 1h Replying to @deepseek_ai New Pricing V4 Flash ≈ GPT-5.4 Nano, but is more expensive than GPT-5.6 Luna / Luna Pro. V4 Pro ≈ GPT-5.4 Mini. ->If you're using V4 Flash, you'd be better off switching to GPT-5.6 Luna or Luna Pro. If you're using V4 Pro, you may not be completely satisfied with it. 5 22 5,051 @valuefp 18m Replying to @deepseek_ai DeepSeekの新価格、以前と比較して最大約12倍。8/16より。まだ中華AIでは最安水準は変わらないが、以前のようなタダ同然とはいかず。 実は一番影響が出るキャッシュヒットの部分でPro、Flashとも 最も値上げしてるのがポイント。ピークでも最大5倍程度か、と思っていると想定を越えるので注意を。 DeepSeekが安く使えていたのは、出力や入力でばなくキャッシュヒットが極端に安かったからです。 328 @srikat 1h Replying to @deepseek_ai For those in Australia AEST time zone: > So for a 9 AM–6 PM AEST workday, you get 5 hours of peak pricing and 4 hours of off-peak pricing. 1 553 @stefdelev 37m Replying to @deepseek_ai lol everyone complaining but no one pushing for them to have compute 🤷🏾‍♂️ 2 465 1h Replying to @deepseek_ai Holy cow, the price is so much higher! I'd rather stick to my ChatGPT. 1 5 2,178 1h Replying to @deepseek_ai I used to have a system that categorized tens of thousands of articles. Unfortunately, the current pricing no longer makes it viable.. 🫤 1 9 3,528 @OdailyChina 41m Replying to @deepseek_ai 🫡 OdailyNews @OdailyChina 42m 是用回GPT還是繼續DS？Kimi價效比也高起來了 梁聖這波直接變牢梁了😂😂 DeepSeek-V4正式版大漲價：Pro高峰輸出暴漲4.5倍 17號正式實行 DeepSee宣佈正式釋出DeepSeek-V4-Pro（版本標識V4-Pro-0813）。這是該公司自4月推出V4系列預覽版以來的生產級正式版本，標誌著V4-Pro從預覽階段全面進入公開可用狀態。 DeepSeek API 的峰時時段為： 09:00 – 12:00（UTC 01:00–04:00） 14:00 – 18:00（UTC 06:00–10:00） V4-Pro 高峰時段每百萬 Token 快取未命中輸入從目前的 3 元漲到 9 元，輸出從 6 元漲到 27 元，分別漲到 3 倍和 4.5 倍。 閒時雖然只有高峰價的一半，但也比現在貴。V4-Pro 輸入為 4.5 元，輸出 13.5 元；V4-Flash 則為輸入 1.5 元、輸出 4.5 元。 🆚和它基本同檔的 GPT-5.6 Luna 剛被 OpenAI @OpenAI 降價 80%，現在輸入只要 1.35 元輸入、8.1 元輸出。 1 769 @Michaelzsguo 1h Replying to @deepseek_ai cache-hit pricing for V4 Pro during peak hours is jumping 12x!! Michael Guo @Michaelzsguo 1h DeepSeek V4 API pricing is now official. As expected, most prices are increasing by roughly 2–4x, with one major exception: cache-hit pricing for V4 Pro during peak hours is jumping 12x. 5 1,103 @howmanysmaII 1h Replying to @deepseek_ai Fumbled the lead award 1 17 1,536 @toi500 1h Replying to @deepseek_ai Off-Peak... 2 11 792 1h Replying to @deepseek_ai This pricing is still good I’ll use the newer model for my tasks for peak quality 491 54m Replying to @deepseek_ai 10 998 43m Replying to @deepseek_ai 漲得太離譜了，樑子。 1 233 @chenmoolin 4m Replying to @deepseek_ai @grok deepseekV4pro 的價格和 grok4.6 相比哪個更便宜？ 1 16 @fajarsentosabuz 1h Replying to @deepseek_ai That 50% off-peak gap could make inference scheduling part of product design. Is DeepSeek planning to publish queue or latency guarantees for teams moving batch jobs into the cheaper window? 154 Load more