Loading...
Loading...
DeepSeek’s permanent 75% price cut for its V4-Pro/code models has created market ripples across China’s AI/ML code-model landscape. The move makes DeepSeek dramatically cheaper per token than rivals like Qwen, GLM, MiniMax and even large models from Google, prompting debate over sustainability and strategic discounting. Observers note DeepSeek is also expanding teams to compete directly with Claude Code and Codex, suggesting product and capacity ambitions beyond a price war. Incumbents may retain high-end customers, but smaller developers and cost-sensitive teams are rapidly shifting. The broader trend highlights intensified competition as startups use aggressive pricing and open-source focus to capture developer mindshare.
DeepSeek's sharp price cuts and investment in a code-focused team change cost-performance expectations for code models and pressure competitors. Tech professionals should reassess procurement, MLOps budgets, and integration plans for cheaper high-quality models.
Dossier last updated: 2026-05-23 03:11:00
@9hills: 为啥Qwen3.7-max 定价那么贵，DeepSeek 定价那么便宜。 最核心的是Qwen 定价里是包含很大一部分折扣空间的，你可以找阿里云的销售谈折扣，这部分就是灵活的。 据我了解，大客户折扣空间较大，但普通开发者以及ToC小龙虾客
借着 DEEPSEEK 这波永久 25 折的机会，我们来回顾一下中国的顶级开源模型现状 比较一下他们的价格/功能/适用场景吧 价格全景（input / output, per M tokens）： - Qwen3-Coder-Next：$0.11 / $0.80 - MiniMax M2.7：$0.30 / $1.20 - DeepSeek V4-Pro：$0.435 / $0.87（降价 75% 后） - Kimi K2.6：$0.95 / $4.00 - GLM-5.1：$0.98 / $3.08
梁圣这次让 DeepSeek V4 系列永久降价，其实最难受的并非 Claude 或者 GPT，甚至可以说对这两家的影响微乎其微，要用这两家的还是会用。 反而受到影响的是国内几家，比如Qwen、GLM、MiniMax 等。智商没多大差距，DeepSeek 还量大管饱。 当然这个优势看能延续多久。如果梁圣还像去年那种��奏去迭代，估计只能压1、2个月。 不过我认为不可能。毕竟现在梁圣融资了，有足够的市场来的钱去冲了😆 有 DeepSeek 最价值标的是用户的福音啊。 我现在就一个问题：梁圣什么时候把多模态也灰度给我。那么多用户都用了一个月了。🤡
@Balder13946731: Deepseek正式将价格打二五折，哪怕是��在最没用的Gemini，价格也是Deepseek的30倍😭😭 Deepseek的优化实在太可怕了。
最近关于DeepSeek的两个消息结合实测，不难得出，DeepSeek有耐心、有能力，天花板还远远没达到： 消息1：DeepSeek正在组建直接对标Claude Code和Codex的Harness团队 消息2：DeepSeek宣布新模型持续25折 实测：DeepSeek最新模型单位质量的价格是$0.0037，相比之下Sonnet4.6是$0.17，Gemini 3.5 Flash到了$0.3，DS 4 Pro在特定任务下质量已和Sonnet4.6相差无几
Tencent Music announced it will forgo exclusive audio copyright rights in response to antitrust conditions tied to its acquisition of Ximalaya, a move mandated by China’s market regulator. SAIC Motor said it will hit a cumulative delivery milestone of 100 million vehicles by month-end, the first Chinese automaker to do so. AI startup DeepSeek cut its V4-Pro API price permanently to one-quarter of the original rate after a limited-time 75% discount and is pursuing a RMB 70 billion financing while pledging to focus on open-source models. WeChat clarified its 2-minute message recall limit is intended to prevent casual tampering of chat records. Other items cover regulatory action on cross-border brokerages, chip and phone launches, and major industry exec comments on AI infrastructure spending.
Show HN: My dad is a forensic accountant. I automated ~62% of his job
糟了， Codex 半夜又偷偷刷新了周额度