跳到正文
只看模型
今天11月8日周日2 条
11月7日周六
  4. Arena.ai45

    在 Code Arena: WebDev 中，@thinkymachines 的 Inkling 2 Small 位列数据与分析类第 5 名！ 在所有开源模型中，它在该类别将排名第 1。相比其前代 Inkling-Small（第 92 名），该类别提升了 +220 分。 总体而言，Inkling 2 Small 在 Code Arena: WebDev 中相比 Inkling-Small 提升了 +240 分。它在所有类别中都展现出大幅进步。 再次祝贺 @thinkymachines 团队！

    引用Arena.ai@arena

    ICYMI: Once its weights are released, Inkling 2 Small from @thinkymachines would be the #1 US open model. With 1648 pts, it’s #8 overall among labs in the Code Arena: WebDev! Inkling 2 Small is available now for fine-tuning and inference on Tinker, with full weights coming soon. Congrats again to the @thinkymachines team!

  8. Arena.ai62

    Arena 公布 thinkymachines 的 Inkling 2 Small 进入 Code Arena: WebDev 榜单，以 1648 分在实验室中排名第 8，权重发布后将成为美国开源模型第 1。该模型现已在 Tinker 上开放微调和推理，完整权重即将发布；此前引用帖中 1654 分版本显示其领先第 2 名 Meta Muse Spark Open 128 分，全球开源模型中排第 4。

    引用Arena.ai@arena

    Exciting news: Inkling 2 Small by @thinkymachines has landed in Code Arena: WebDev with 1654 pts (AutoEval). Once its weights are released, it would rank #1 among US open models! This is +128 pts ahead of Meta’s Muse Spark Open at #2 (US open). Inkling 2 Small would rank #4 among open models globally, and ranks #20 overall. This is +245 pt improvement over its previous variant, Inkling Small (#32 open with 1409 pts). This model is priced at $1.00/1M output and $0.30/1M input. More info coming soon on its median cost per task in the Code Arena: WebDev, and results in the Agent Arena. Congrats to the team at @thinkymachines on this release!

11月6日周五
  1. Rohan Paul53

    Nace AI 发布 Drex 开源决策模型技术报告，作者评价其用 27B 模型只做选择不做生成，专攻 Agent 中的大量小决策。Drex Max 在 Decision Bench v4 上与 GPT-6 Sol 统计持平，单次前向传播即输出各选项概率；成本 $0.06/1K 决策、延迟 183 ms，对比 Claude Opus 5.5 的 $10.17 和约 3.5 秒。

    引用Nace AI@NaceAI

    How are Decision Models Trained? Today, we are publishing the Drex Technical Report for the developer community! 🚀 Discover details on how the State-of-the-art Open-Source decision models are trained with RLAF. Build more at: console.nace.ai (try for free) P.S. Drex Max 1.1 and Drex for Finance are closed sourced and designed for enterprise use-cases

  3. TechCrunch：AI55

    Underdog 发布 32MB 语音模型 Woof 并推出浏览器演示游戏

    Underdog 发布仅 32MB、3100 万参数的语音生成模型 Woof，并推出一个复古科技主题的浏览器游戏演示，可在无网络连接的多数设备上运行，模型开放下载。创始人 Sigil Wen 表示目标是训练开源、美国制造且支持数据私有的模型，此前公司还推出过 167MB 的语音转文本模型 Bark 和一个可跑在笔记本上的 270 亿参数模型。

  4. Tencent Hy42

    设计基准 👇

    引用OpenDesign Developers@OpenDesignDevs

    Hy4 by @TencentHunyuan scores just 0.1 points behind GPT-6.1 Sol in our design benchmark, at 41% lower estimated cost. We tested 23 models on real-world design tasks based on user requests. Hy4 ranks #4. Now on OpenDesign. Full results ↘️

  6. 公众号：蚂蚁百灵66

    蚂蚁百灵发布 Ling-3.2-flash：原生多模态，560B 总参数单步激活 25B

    蚂蚁百灵发布原生多模态模型 Ling-3.2-flash，总参数 560B、单步激活 25B，支持 1M 原生长上下文，重点强化视觉反馈自检与异常自适应恢复能力。

    推荐理由：原文给出 Ling-3.2-flash 的参数规模、视觉自检与异常恢复机制和推理加速数据，可据此判断其在长程任务中的可用性。

  7. Tencent Hy49

    #1 open-weights model on APEX-Agents 🙌

    引用Mercor@mercor

    Hy4 from @TencentHunyuan is the new #1 open-weights model on APEX-Agents. 77.6% Pass@1 (#3 of 69) 85.1% mean score (#3 of 69) Here’s the updated APEX-Agents open-weights top 5 (Pass@1): 1. Hy4: 77.6% 2. DeepSeek-V4.1-Flash-1017: 69.8% 3. Kimi K3.1 Pro: 65.5% 4. Qwen3.8-Max: 63.3% 5. MiMo V2.6 Flash RL: 57.4% Hy 4 is also the top open-weights model in every APEX-Agents domain: Management Consultant: 85.6% Investment Banking Analyst: 75.3% Corporate Lawyer: 71.9% Congrats to @TencentHunyuan. See the full leaderboard: mercor.com/apex/

  9. IT之家·人工智能43

    webAI 开源 TwIL-8B 推理模型：聚焦数学与编程，长板表现优于 Gemma 4 31B

    webAI 开源了基于 IBM Granite-4.2-8B 的 TwIL-8B 推理模型，针对数学、代码和指令跟踪优化，采用非商业许可。该模型在 MATH-500、AIME 2025 两项数学基准上取得参比模型最佳成绩，强于尺寸数倍的 Gemma 4 31B，LiveCodeBench 同为同组最佳，IFBench 与 Gemma 4 12B 持平。

  10. Tencent Hy43

    感谢 @arena！🎉 很高兴看到 Hy4 成为 Agent Arena 上排名第一的中国模型。

    引用Arena.ai@arena

    Hy4 by @TencentHunyuan is now officially the #1 best Chinese model in Agent Arena! With over 12K+ real-world, long-horizon agentic sessions from our global community of users, Hy4 has also placed Tencent Hunyuan as the only Chinese lab in the top 5. It is now showing +5.4% net improvement, and sits firmly on the Agent Arena Pareto frontier at a $0.57 median cost per task. Congrats again to the @TencentHunyuan team!

  17. Arena.ai60

    Arena 宣布腾讯混元 Hy4 在 Agent Arena 榜单成为排名第一的中国模型，位列总榜第 5，也是前 5 中唯一的中国实验室模型。基于超过 1.2 万条真实长程智能体会话，Hy4 净改进达 +5.4%，进入 Pareto 前沿，中位成本为每任务 $0.57。引用内容补充称 Hy4 从 preview 的第 25 名（+1.1%）提升至第 14 名（+4.7%），在用户确认成功、情绪反馈、可操控性和 Bash 可靠性等信号上均有提升。

    引用Arena.ai@arena

    Exciting news from Agent Arena: Hy4 by @TencentHunyuan just landed #14 overall. Currently proprietary, but once its weights are released it will tie with Kimi K3.1 Pro for #1 among open models! Across +8.5K real-world, long-horizon agentic sessions from our global community of users, Hy4 delivers +4.7% net improvement and joins the Agent Arena Pareto frontier. At $0.55/task, Hy4 costs about 62% less than Kimi K3.1 Pro (Max) at $1.46/task, and the two models’ scores are within 1 percentage point. This release is a +3.6 percentage pt improvement from Hy4 preview at #25 with +1.1%. In fact, Hy4 improved across all key signals: - Confirmed Success (explicit “yes, that worked” from users): #11 with +8.7% (up from #16 with +4.9%) - Praise vs Complaint (sentiment in user reactions): #18 with +3.8% (up from #24 with +1%) - Steerability (responding to user corrections): #14 with +5.7% (up from #27 with +0.21%) - Bash Reliability (successful shell commands): #20 with +6.2% (up from #41 with -3.1%) Congrats to the @TencentHunyuan team on this release!

  20. Arena.ai63

    Arena 宣布 Thinking Machines 的 Inkling 2 Small 进入 Code Arena: WebDev，AutoEval 得 1654 分，约列总榜第 20、美国开源模型第一。

    引用Thinking Machines@thinkymachines

    Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor. Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon. thinkingmachines.ai/news/inkl…

  21. Karina56

    Karina Nguyen 分享一个 Inkling 2 Small 智能体的实测轨迹：当申报工具只返回 PSEG 的目录时，该智能体自行下载完整的 3.7MB 10-Q，用 Python 解析，并发现需从经营现金流中剔除的 12 亿美元抵押品释放，得分 0.82，高于次优模型的 0.67。引用的 @thoughtfullab 内容称其在 DiligenceBench 上表现良好，该模型为 594B 参数、17B 激活参数的 MoE，定价 $0.30/M 输入和 $1/M 输出 token。

    引用Thoughtful@thoughtfullab

    Inkling 2 Small can generate a full public equity diligence memo for just $0.69 using our finance harness. We got early access to the model from @thinkymachines and have been evaluating it on DiligenceBench, where it’s performing remarkably well. It’s a 594B-parameter MoE model with 17B active parameters, priced at $0.30/M input and $1/M output tokens.

  23. Thinking Machines Lab：News79

    Thinking Machines Lab 发布 Inkling 2 Small：594B 参数 MoE 模型，下周开放 Apache 2.0 权重

    Thinking Machines Lab 发布第二代模型 Inkling 2 Small，为 594B 总参数、17B 激活的 MoE 模型，支持图像输入与 variable effort，Artificial Analysis Intelligence Index 得分 31。

    推荐理由：官方完整披露训练配方、RL 规模与安全评测，开放权重计划和基于基准的定位值得跟踪其后续表现。

  26. Mira Murati54

    Mira Murati 宣布 Inkling 2 Small 即日起可在 Tinker 上进行微调和推理，完整权重即将开放。据引用的 Thinking Machines 官方内容，该模型在 AA Intelligence Index 上得 31 分，较前代大幅提升。

    引用Thinking Machines@thinkymachines

    Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor. Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon. thinkingmachines.ai/news/inkl…