图片由AI生成 凭借成功预测Polymarket筛选题目，连续登顶FutureX全球榜首的MiroMind团队，正式发布了其自研旗 舰搜索智能体模型MiroThinker 1.5。 MiroMind由全球知名创新企业家、慈善家陈天桥，与清华大学知名AI青年学者代季峰教授联合发起。 去年，陈天桥提出，发现式智能才是真正意义上的通用人工智能这一重磅创新理念，引发全球业内人士 关注。他同时提出，建设发现式智能的5种关键能力，其中一项能力，是在未知条件下重建对世界的理 解，这正是MiroMind的使命。 在过去7个月里，MiroMind在思考一个更本质的问题：智能的奇点究竟在哪里？ 他们给出的答案不是把世界背进参数里，而是押注"发现式智能"：真正的智能不靠全知，而靠会研究、 会查证、会修正——像顶级情报官一样对外极速取证、对内严苛去伪存真；像严谨研究员一样在不确定 性里逼近真相，最终把预测未来从特权变成能力。 MiroThinker 1.5 ：30B参数，闯入全球搜索智能第一梯队 MiroMind团队在AGI竞技场上，不信奉"大力出奇迹"，而是追求以高智效比为核心的巧劲。 MiroThinker-v1.5-30B仅用 ...

凭借成功预测 Polymarket 题目，连续登顶 Future X 全球榜首的 MiroMind 团队，于今日（1 月 5 日）正式发布其自研旗舰搜索智能体模型 MiroThinker 1.5。 MiroMind 由全球知名创新企业家、慈善家陈天桥，与清华大学知名 AI 青年学者代季峰教授联合发起。去年陈天桥提出发现式智能才是真正意义上的通用 人工智能这一重磅创新理念，引发全球业内人士关注。他同时提出建设发现式智能的 5 种关键能力，其中一项能力是在未知条件下重建对世界的理解，这 正是 MiroMind 的使命。 BrowseComp 性能对比 | Model | Param | Browse | Browse | HLE | GAIA | | --- | --- | --- | --- | --- | --- | | | | Comp | Comp-ZH | | Val-165 | | Closed-Source Models | | | | | | | GPT-5-High | l | 54.9 | 63.0 | 41.7 | 76.7 | | ChatGPT Agent | l | 68.9 | - ...

鹭羽 发自 凹非寺 量子位 | 公众号 QbitAI 新年刚至，陈天桥携手代季峰率先打响开源大模型 的第一枪。 正式发布其自研的旗舰版搜索智能体模型—— MiroThinker 1.5 ，堪称智能体模型领域的最强小钢炮。 最直观的还是基准测试上的性能评测： 在面对GPT-5-High、Gemini-3-Pro、DeepSeek-V3.2等一系列国内外顶尖模型，MiroThinker 1.5在四项基准测试中的表现都毫不逊色： HLE-Text （人类终极测试） ：39.2% BrowseComp （网页检索类大模型基准测试） ：69.8% BrowseComp-ZH （BrowseComp的中文适配版本） ：71.5% GAIA-Val-165 （GAIA基准测试验证集） ：80.8% | Model | Param | Browse | Browse | HLE | GAIA | | --- | --- | --- | --- | --- | --- | | | | Comp | Comp-ZH | | Val-165 | | Closed-Source Models | | | | | | | GPT-5- ...

Core Viewpoint - MiroMind team has launched its flagship search intelligence model MiroThinker 1.5, emphasizing the concept of "discovery intelligence" as a path to true general artificial intelligence, focusing on external information interaction rather than merely increasing internal parameters [1][10]. Group 1: Model Performance and Comparison - MiroThinker 1.5-30B achieved performance comparable to many 1 trillion parameter models while using only 1/30 of the parameter scale [4]. - In key benchmark tests, MiroThinker 1.5-235B ranked among the top globally, demonstrating its effectiveness despite a smaller parameter size [4]. - MiroThinker 1.5-30B exhibited a significantly lower inference cost of $0.07 per call, which is only 1/20 of the cost of Kimi-K2-Thinking, while also providing faster inference [9]. Group 2: Interactive Scaling and Training Mechanism - MiroMind team has shifted from traditional scaling laws focused on internal parameter expansion to "Interactive Scaling," which emphasizes external information interaction to enhance model performance [10][12]. - The training process encourages models to engage in evidence-seeking behaviors, breaking down key judgments into verifiable sub-hypotheses and actively querying external data [19]. - The model is trained under strict temporal visibility constraints, ensuring it learns to make judgments based only on past information, thus avoiding future leakage [17][20]. Group 3: Unique Training Approaches - MiroThinker 1.5 employs a "scientist mode" rather than a "test-taker mode," focusing on verification and correction rather than memorization [11]. - The model's training paradigm includes a time-sensitive training sandbox, which forces it to operate under real-world conditions of incomplete information and noise [18]. - The training emphasizes iterative verification and self-correction, allowing the model to adjust its hypotheses based on conflicting evidence [19]. Group 4: Market Predictions and Applications - MiroMind has demonstrated its predictive capabilities in stock market scenarios, accurately identifying stocks with high potential for upward movement amidst market noise [22][25][30]. - The model is also applied to predict significant events that may impact major companies, providing insights into potential market reactions and volatility [31].