Keep on to blur preview images; turn off to show them clearly
DeepSeek V3.2正式版刚刚出来了，强化了Agent能力，融入了思考推理 不同于过往版本在思考模式下无法调用工具，V3.2是个【将思考融入工具使用】的模型，并且同时支持思考模式与非思考模式的工具调用 两个模型，DeepSeek-V3.2和DeepSeek-V3.2-Speciale V3.2以平衡推理能力与输出长度为主，适合日常使用。比如，问答和通用Agent任务场景 在Benchmark上，达到GPT-5水平，略低于Gemini-3.0-Pro，比Kimi-K2-Thinking其输出长度大幅降低，减少了计算开销与等待时间 Speciale以探索模型能力边界为主，是V3.2的长思考增强版，同时结合了DeepSeek-Math-V2的定理证明能力 在主流推理基准测试上的性能表现媲美Gemini-3.0-Pro 并斩获IMO 2025、CMO 2025、ICPC World Finals 2025及IOI 2025金牌 在高度复杂任务上，Speciale大幅优于标准版本，但消耗的Tokens也多，成本更高 目前仅供研究使用，不支持工具调用，暂未针对日常对话与写作任务进行专项优化 V3.2的思考模式也增加了对Claude Code的支持，可以通过将模型名改为deepseek-reasoner，或在Claude Code CLI中按Tab键开启思考模式进行使用 但需要注意的是，思考模式未充分适配Cline、RooCode 等使用非标准工具调用的组件，建议在使用此类组件时继续使用非思考模式 官方网页端、App、API均可用正式版DeepSeek-V3.2 #DeepSeekV32 #DeepSeek
DeepSeek-V3.2 HuggingFace:https://t.co/qai0089WTX ModelScope:https://t.co/I5CjK7pvwH DeepSeek-V3.2-Speciale HuggingFace:https://t.co/ydRWjqOO9L ModelScope:https://t.co/ctwA7LHaFU API文档：https://t.co/lLXQDA92Xl
enough preamble. The most important part in every Whale paper, as I've said so many times over these years, is “Conclusion, Limitation, and Future Work”. They say: Frontier has no knowledge advantage. Compute is the only serious differentiator left. Time to get more GPUs.
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
disappointingly, the most efficient and scalable way to manage limited context of a search agent is to just yeet all tool calls. Summarizing does very poorly. I think it's a more “traumatic” context trajectory alteration than lacunae, and has to be trained for specifically.
enough preamble. The most important part in every Whale paper, as I've said so many times over these years, is “Conclusion, Limitation, and Future Work”. They say: Frontier has no knowledge advantage. Compute is the only serious differentiator left. Time to get more GPUs.
I'm really happy about the first comment. But I'm even happier about the response. People now think that AI-generated actors/videos made with VIDEO AI ME are real. That means I'm doing my job well. VIDEO AI ME passes the Turing Test for video quality with success 😊
I just launched a new startup today → VIDEO AI ME → Create your own AI clone and make studio-quality videos that go viral → https://t.co/eU2hD1uyoV
AI 养蛊委员会之 Token 在燃烧 LLM Council 把你所有能找到的 AI 攒个局，什么 GPT-5.1、Gemini 3.0、Claude 4.5 全拉进一个群里，把名字都盖住，让它们互相审查、互相比稿，还得排个名次出来。 让你手里的 AI 们内卷起来 ，最后获得大力出奇迹的终极缝合怪答案。 https://t.co/SWRlJM7Ozm
🧠在家居士 | 🥦素食者 | 🏃🏻马拉松爱好者 | 💰省钱小能手 | 搭🪜技术资深学者 | 👨💻科技宅 | 🆕更新狂 | 🆅 六边型战五渣
Austin - Workers Tech Talks ATX is this week! Join us for a night of pizza and deep dives into building on the Workers platform 🤠 RSVP below! 💫 https://t.co/zVaotEQoNJ
Have questions, or building something cool with Cloudflare's Developer products? We're here to help. For help with your account please try @CloudflareHelp