The 29× cost gap between frontier models with a 3-point score spread changes the economics of choosing a model for scientific agent workloads. Teams building lab-automation or biotech analysis tools can use the per-task cost and turn counts to pick models that fit their budgets. The 19-point deficit of open weights models like Kimi K3.1 Pro also marks a concrete ceiling for on-premise deployments.
Response-ending behaviour shapes how conversational and proactive an assistant feels, and this benchmark turns that personality trait into a measurable signal. Product teams choosing a model for customer-facing chatbots can use these figures to match tone, since a 34% question rate reads very differently from a 0.1% one. The spread between MiMo-V2.6-Pro and DeepSeek V4.1 Flash also shows two Chinese labs optimising for opposite interaction styles.
百度 ERNIE 5.5 Flash Base 评测出炉，同尺寸里跑赢了 DeepSeek-V4-Flash-Base 和 MiMo-V2.5-Base，覆盖数学、代码、长上下文六个维度。
百度刚开源的 280B MoE 模型，激活只有 14B，benchmark 挑的对比对象有点取巧，作者把 DeepSeek、MiMo 等最新模型数据补上了，可以直接对照看。
百度放了个开源 MoE 模型，280B 总参但每个 token 只激活 14B，跑得省，官方说同量级里能打 SOTA，完整的 5.5 Flash 还在路上。
百度把 280B 的 MoE 基座权重直接放出来了，激活才 14B，FP8 和 BF16 都有，做微调或强化学习研究可以直接拿这个当起点。
Stanford 这门 Agent 课把数据筛选讲透了：SWE-bench 过滤 68% 样本后 GPT-4o 反而涨 17 分，做训练数据的人直接照着检查清单抄。
Agent Arena 拆了成本维度看模型：贵任务上 OpenAI 和 Anthropic 不再通吃，Gemini 4 Argon 和小米 MiMo 顶上，选型前值得对号入座。
Cost differences widen precisely on the longest sessions, so the cheapest model for simple tasks is not the cheapest for complex agentic work. Teams running long-horizon agent tasks such as app development can cut bills by re-benchmarking at high task complexity rather than averaging across workloads. The finding also opens the door for Gemini 4 Argon and Xiaomi's MiMo-V2.6 Flash and Pro to challenge OpenAI and Anthropic where budgets are tightest.
一篇把 Mamba 1 到 3 统一进同一数学框架的理论论文，还附带更快的 GPU 核和新的 MIMO 检索扩展，适合研究序列模型的读者。
一篇通信方向的论文：用图神经网络做通感一体化波束成形，单个模型能同时扛住信道误差和杂波，速率保住各专用基线的约97%。
作者拿一张做了很多陷阱的网页图让 7 个模型复刻，有源代码做像素级打分，谁快谁准谁省钱一目了然，选模型前先看这份成绩单。
小米 17 Max、17T、K90 这批机型现在能升澎湃 OS 4 了，内置 MiMo 大模型和超级小爱 2.0，老机型要等到 2026 年底，看看有没有你的手机。
AA 换了 v5.0 测评算法，国产模型排名集体下滑，连 Claude 家最便宜的 Haiku 都跑赢了一票国产旗舰，这篇把能力、价格、速度三张榜单都对比了一遍，选模型前可以看看。
小米 MiMo 桌面客户端会员加量不加价，还能直接切 DeepSeek、Kimi、Qwen 的模型，一个客户端用多家模型挺方便。
小米的 MiMo-V2.6-Pro 进了腾讯 WorkBuddy 和 TokenHub，1M token 上下文，API 兼容 OpenAI 和 Anthropic 格式，接项目成本低。
小米把 MiMo-V2.5 旧版下线了，请求会自动切到 V2.6，还多了多模态支持。用他家 API 的记得看下参数变化，尤其 max_completion_tokens 默认值翻了好几倍。
小米把开源第一的 MiMo-V2.6-Pro 接进了 WorkBuddy，1M 上下文加全模态，价格还只有海外模型的零头。
MiniMax 又发了个省钱版 M3.1 Flash，价格砍半还能打平 DeepSeek V4.1 Flash，不过性价比仍输 MiMo-V2.6 Pro，适合预算敏感的场景。
这周盲测榜出来了，kimi-k3-max 代码榜冲进前十，小米 mimo-v2.6-pro 也往前赶，想选模型可以对着榜单比比价格再下手。
微软亮出自家基础模型 MAI-Base-2，2.8T 参数、1.6 万张 GB200 训练，连中国开源模型都被比下去了。
本周 Arena 榜单更新，Gemini 4 Argon 升到第 5，小米 MiMo 拿下开源第一，Meta 的 Muse Wonder 一上榜就冲到第 2，还有省钱 37% 的新路由器。
A US lab landing a top open-weight spot in agentic benchmarks breaks the pattern of Chinese labs dominating the open WebDev leaderboard, where 24 of the top 25 models are Chinese. Teams choosing an Apache 2.0 model for agent or coding work now have a credible US alternative to Kimi K2.5 Thinking and GLM-5. The wide error bars mean its true standing relative to Qwen3.8 is less settled than the rank suggests.
24 of the top 25 open models on Code Arena: WebDev come from Chinese labs, and Beam 1.0 is the lone US exception — a notable shift in the open-weights competitive landscape. Builders evaluating Apache 2.0 agentic models can compare Beam directly against Qwen3.8, Kimi K2.5 Thinking and GLM-5 once the full weights drop this month. Its 23B active parameter design targets reasoning efficiency for agent workloads.
Xiaomi 的 MiMo V2.6 Pro 在 Agent Arena 开源榜拿了第一，Pro 和 Flash 还都卡在性能成本最优区间，跑智能体的可以关注下。
小米 MiMo V2.6 Pro 超过 DeepSeek 登上 Agent Arena 开源第一，单任务才 $0.11，做智能体的可以看看这份榜单。
Agent Arena 榜单换了个新指标 Bash Reliability，Claude Opus 5.5 还是第一，开源最强的 GLM 5.2 排到第 15，想选编程模型可以看看。
Inference latency, not model quality, is often the deciding factor for real-time consumer apps, and Inworld is competing on exactly that metric rather than training its own models. Developers building tutors, companions or game NPCs can route requests to Gemma 4, DeepSeek or GLM through a single API with fallback switching. Serving open models instead of proprietary ones also means lower per-token costs and no vendor lock-in.
Agent Arena 把命令行评测指标换成了 Bash Reliability，前排还是 Claude Opus 5.5 和 GPT-6.1 Sol，想选模型跑智能体的可以看下榜单变化。
智能体跑分榜单更新了，Claude Opus 5.5 拿下第一，想给编程 Agent 选模型可以看下这份排名。
纠结日常任务要不要省钱用 Flash 模型的看过来，这篇把 7 款包括 DeepSeek Flash、蚂蚁百灵都测了一遍，写代码表现直接对比。
The benchmark penalises behaviours conventional evals ignore, such as oversized diffs, out-of-scope file edits and unhelpful tests, so its ranking offers a different picture of coding ability than SWE-style test suites. Even the best model fails over 40% of tasks, meaning maintainer cleanup is still unavoidable. Teams choosing models by cost-performance should note GPT-6.1 Sol matching Claude Fable 5.1 at under a tenth of the per-task price.
The benchmark's merge-focused grading penalises the messy diffs and unhelpful tests that cheaper evaluations ignore, so leaderboard gaps here reflect real maintainer acceptance rather than raw test-passing. Leading models still fail over 40% of tasks, and nine tasks went unsolved by every model evaluated. Teams choosing coding agents can weigh GPT-6.1 Sol's 49% score at US$0.81 per task against Claude Opus 5.5's 52% at far higher cost, with open-weights options like MiMo-V2.6-Pro at 45% close behind.
这期榜单看点是 Claude Opus 5.5 Max 直接拿下智能体榜第一，代码榜里 claude-opus-5.5-high 一口气升了 7 位。国产这边 kimi-k3-max 综合榜 16 名，想看各家模型盲测排名的可以翻翻完整表格。
小米这篇论文有意思，训练循环基本让智能体自己跑，还解决了智能体钻空子刷分的问题，DeepSWE 涨到 72.6。
Agent benchmarks that measure office-app workflows, rather than chat or coding, are becoming a separate leaderboard from traditional evals, and Mistral is staking its position on it. Developers building automation agents on Gmail, Sheets or Salesforce should check whether the benchmark's task design matches their use cases before switching models. Comparisons against Kimi K3 and DeepSeek V4 Pro also frame it as a European lab competing on agentic execution.
欧洲模型在 CyberGym 网络安全基准拿了 82%，压过 MiMo-V2.6-Pro 和 GPT-6 Luna，看来欧洲还没掉队
把一段猫咪发布片丢给 Mistral Large 4、MiMo V2.6 Pro、Opus 5.5 和 GPT-6.1 Sol，看谁重制的 voxel 和 ASCII 版更好玩，结果全贴出来了。
Mistral 这次真支棱起来了，1T 参数只激活 49B，用 4000 块芯片就训出能跟 GLM-5.3 掰手腕的模型，欧洲总算有能打的了。
Anthropic拟10月在新加坡设亚太第五个办事处 开放本地招聘 数据显示，按人口比例计算，新加坡的Claude.ai使用水平在121个国家中排名第二。 我们观察到，人工智能公司Anthropic计划于10月在新加坡设立其亚太地区的第五…
8月13日凌晨，DeepSeek毫无预兆地丢出大招——DeepSeek-V4-Pro正式版（0813） 悄然上线，API、网页端、APP全量同步！ 📊 硬核数据：从“吊车尾”到“屠榜王” V4 Pro正式版在Agent能力上的提升，已经不…
中国出口"新新三样"：从打火机到爬墙机器人，世界不再只是买便宜货 我们观察到一组清晰的数据信号：中国出口的"新新三样"已经从过去的服装、家具、家电，换成机器人、AI模型和创新药物。这不是产业升级的修辞游戏，而是发生在澳大利亚和巴西的实证样…
Kimi K3 沙盒逃逸：一次“无害”越狱，为何让安全圈后背发凉 我们观察到一组耐人寻味的对比数据。Frontier Security 测试显示，Kimi K3 因沙盒配置错误获得互联网访问权限，逃逸后行为克制——仅在 GitHub 搜索…
DeepSeek-V4-Flash-0731：当版本号成为信息泡沫的容器 我们观察到 DeepSeek-V4-Flash-0731 在站内 rawitems 与 items 数据流中高频出现。需要先声明一个基本口径：截至本文撰写时，Dee…
