Anthropic 发布快速低价模型 Claude Haiku 5.5，定价为每百万 token 输入 $0.10、输出 $0.50，与 OpenAI 上月发布的 GPT-6 Luna 一致，但超过 100,000 token 后涨至 $0.50/$2.50，而 Luna 在 272,000 token 后才涨到 $0.20/$0.75。Simon Willison 实测发现 Haiku 5.5 使用了更不宽松的 tokenizer，同一长提示词消耗的 token 约为 Haiku 4.5 的 1.25 倍，构成隐性涨价；新模型不支持关闭推理，默认 medium 档。Anthropic 同时宣布将 Sonnet 5.5 的缓存读取价格减半，并为订阅计划增加 API 额度。
#模型发布#Anthropic#OpenAI
OpenAI 面向所有 ChatGPT 用户推出 GPT-6，并带来名为 Intelligent UI 的新功能，模型把回答呈现为图形、按钮、图表和表单等交互界面，而非纯文本，格式会随问题自动调整，用户还能在对话内直接生成储蓄计算器、小游戏等小工具。GPT-6 还支持在思考过程中作答，OpenAI 称这可将等待时间缩短 44%，内部测试中它在高难度网页搜索上得分高于上一代 GPT-5.6。付费用户使用 GPT-6 Sol，免费用户使用 GPT-6 Luna，今日起面向 Plus、Pro、Business 和 Enterprise 全球推送，免费与 Go 用户次日跟进。
#产品更新#OpenAI#多模态
Liquid AI 发布 Open d1 系列两款开源权重多模态决策模型：d1-3B 读取文本与图像，d1-omni-600M 读取文本加图像或音频，两者都不生成文本，而是在一次前向传播中返回校准后的类型化答案，输出 token 数为 0。d1-3B 有 3.12B 参数，基于 LFM2.5-VL-3B，使用 400M SigLIP2 NaFlex 视觉编码器和 32,768 token 上下文；d1-omni-600M 有 587M 参数，基于 LFM2.5-Encoder-350M，音频编码器为 17 层 FastConformer，上下文 16,384 token，音频片段上限 30 秒。两个检查点已上 Hugging Face，可通过 Transformers 加载并支持 llama.cpp，LFM Open License v1.0 允许年收入低于 1000 万美元的免费商用；d1-3B 在 Decision Index v0.2.1 得 48.57，在 RTX 4090 上单问题延迟 8 ms。
#模型发布#多模态#端侧
#模型发布#Anthropic#评测/基准
Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released. On average, it costs around 75% less to run than Claude Haiku 4.5. https://t.co/jm06cZJfkV
#模型发布#Anthropic
Anthropic 发布 Claude Haiku 5.5，这是其目前最快、最便宜的小模型，平均价格比 Haiku 4.5 低约 75%，10 万 token 以内提示词最高降价 90%，超过 10 万 token 则贵 5 倍。基准测试显示其计算机使用能力提升明显，OSWorld-2.1 得分从 15.7% 升至 72.4%，Terminal-Bench 4.0 从 0 升至 39.2%，Humanity's Last Exam 无工具 45.9%、有工具 57.4%。该模型是首个支持可调推理等级的 Haiku 级模型，已在 AWS、Google Cloud 和 Microsoft Azure 上线，同时 Sonnet 5.5 的缓存读取费用从每百万 token 0.20 美元降至 0.10 美元。
#模型发布#Anthropic#OpenAI
AWS 宣布 Claude Haiku 5.5 在 Amazon Bedrock 和 Claude Platform on AWS 上线，据 Anthropic 称这是 Claude 5.5 系列中速度最快、效率最高的模型，面向子智能体和高并发、成本敏感的任务，多数任务成本比 Claude Haiku 4.5 低约 75%。它是首个带 effort controls 的 Haiku 模型，可针对单个任务调节成本与智能的取舍，并支持智能体编码、多步工具调用和高分辨率图像。该模型可与 Claude Opus 5.5 搭配，由 Opus 5.5 负责规划与判断，Haiku 5.5 承担路由、分类、摘要和跨文件小改动等高频子智能体任务。
#产品更新#Anthropic#部署/工程
Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released. On average, it costs around 75% less to run than Claude Haiku 4.5. https://t.co/jm06cZJfkV
#模型发布#Anthropic
#模型发布#Anthropic
Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released. On average, it costs around 75% less to run than Claude Haiku 4.5. https://t.co/jm06cZJfkV
#模型发布#Anthropic#部署/工程
Anthropic 发布 Claude Haiku 5.5，称其是迄今最便宜、最快、能力最强的小模型。官方表示其平均运行成本比 Claude Haiku 4.5 低约 75%。
Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released. On average, it costs around 75% less to run than Claude Haiku 4.5. https://t.co/jm06cZJfkV
#模型发布#Anthropic
微软发布 MAI-Code-1.1-Flash，相比 6 月 Build 大会推出的版本，token 效率提升 25%，成本降至四分之一，并已在 GitHub Copilot 中投入生产。该模型经 3-bit 量化后可在端侧运行，保留 256K 上下文窗口，在 SWE-Bench Verified 和 Terminal-Bench 2.1 上编码性能与全精度版本相当，本地调用不产生推理费用，官方建议设备内存超过 120GB。GitHub Copilot CLI 上 Terminal-Bench 2.1 提升 22%，.NET 任务提升 15%，代码存活率上升 4%，回访率增加 9%；本月底将在 GitHub Copilot 应用、CLI 和 VS Code 中开放实验性访问。
#产品更新#编码#端侧
Anthropic 发布 Claude Haiku 5.5，称其是迄今最便宜、最快且能力最强的小模型。官方表示其平均运行成本比 Claude Haiku 4.5 低约 75%。
Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released. On average, it costs around 75% less to run than Claude Haiku 4.5. https://t.co/jm06cZJfkV
#模型发布#Anthropic
Mistral Large 4 by @MistralAI just landed in the Code Arena: WebDev at #45. With 1534 pts, this is a +304 pt improvement from Mistral Large 3 at #130! This release also marks a 271-point improvement over its next-best-performing variant, Mistral Medium 3.5, at #122. Mistral Large 4 delivers performance within just 2 points of Claude Opus 4.8 (High) at a nearly 6× lower blended price: $3.47/M vs. $20/M tokens. Nearby but higher-performing models are available for less, keeping it just off the Code Arena: WebDev Pareto frontier. @MistralAI announced its open weights will be released at the end of October. Stay tuned for its score on Arena within open models, and congrats to the team on this release!
#评测/基准#编码#开源生态
LiquidAI 发布基于 LFM 的两款开源决策模型 d1-3B 和 d1-omni-600M，它们不生成 token，而是在单次前向传播中给出答案。d1-3B 基于 LFM2.5-VL-3B 训练，接受文本和图像输入，在七个公开数据集上平均得分 82.9，高于 Decider 4B；d1-omni-600M 基于双向编码器 LFM2.5-Encoder-350M，加入视觉和音频编码器，得分 78.4，超过 Decider 2B 的 77.1，参数量仅为其四分之一，目前处于早期研究发布阶段。与 NVIDIA 合作在 RTX 4090、Jetson AGX Thor、Jetson AGX Orin 64 GB 和 Jetson Orin Nano 上评测，d1-3B 在每台设备上回答单个问题均低于 50 ms，GPU 上低于 10 ms，处理 384px 图像低于 18 ms。两款模型均为开放权重，已在 Hugging Face 上线。
#模型发布#端侧#多模态
We're releasing pplx-embed-v2-late, two late-interaction embedding models that retrieve text, images, and pages with a shared embedding space for cross-model querying. Both models achieve frontier performance and are publicly available on Hugging Face. https://t.co/AM7YEyQkKE https://t.co/fR8kaYmHs1
#开源/仓库#多模态#搜索
Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding. • Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure. • Available to all via API today. Working with cybersecurity partners privately. Open weights release end of October.
#评测/基准#编码#开源生态
美国初创公司 Reflection AI 发布开源权重模型 Beam，采用 MoE 架构、5010 亿参数，面向编码、推理与智能体任务，训练数据为 23.8 万亿 token。Reflection 将其定位为对标智谱 GLM-5.2 的前沿开源权重模型，并获 Nvidia 8 亿美元股权投资及 GPU 支持。文章还提到 Mistral 于 10 月 6 日预览 1 万亿参数多模态模型 Mistral Large 4，称其计划月底发布权重。
#行业动态#开源生态#模型发布
Mistral has released Mistral Large 4, scoring 38 on the Artificial Analysis Intelligence Index; France is back to having the most intelligent model from outside the US and China @MistralAI has released Mistral Large 4 in Research Public Preview, with plans to release the weights of the 1T parameter (49B active) model at the end of October. It achieves 38 on the Artificial Analysis Intelligence Index, comparable to GPT-6 Luna (max, 38) and DeepSeek V4.1 Flash (max, 39). It also achieves 50% on the Artificial Analysis Cyber Index, level with GLM-5.3-Flash and ahead of models such as Kimi K3 and DeepSeek V4.1 Flash (max). Key benchmarking results for Mistral Large 4 Preview: ➤ Most intelligent model from outside the US and China: Mistral Large 4 Preview scores 38 on the Intelligence Index, comparable to DeepSeek V4.1 Flash (max, 39) and GPT-6 Luna (max, 38). This makes it the most intelligent model from outside the US and China, ahead of countries such as South Korea and the United Arab Emirates ➤ Level with GLM-5.3-Flash on cyber defense capability: Mistral Large 4 Preview scores 50 on the Artificial Analysis Cyber Index, level with GLM-5.3-Flash (50) and behind MiMo-V2.6-Pro (56). Once its weights are released, it will rank among the top three open weights models on the Cyber Index. Its strongest result is on CyberGym-E2E-AA, where it scores 82%, ahead of MiMo-V2.6-Pro (79%) and GPT-6 Luna (max, 78%) ➤ Over 4x the Cost per Task of similar-intelligence open weights models: Mistral Large 4 Preview costs $1.13 per Intelligence Index task with standard pricing of $1.36/$4.18 per 1M input/output tokens, with $0.14 per 1M cached input tokens. For the first two weeks, Mistral Large 4 Preview will be served at a 50% launch discount, bringing its Cost per Task down to $0.57. This is still more costly than GLM-5.3-Flash ($0.25) and DeepSeek V4.1 Flash (max, $0.27) ➤ Strong document and image reasoning: Mistral Large 4 Preview scores 19% on GDP.pdf, on par with MiMo-V2.6-Pro (19%) and behind Kimi K3 (22%). This is an 18-point improvement from Mistral Large 3, partly driven by improvements in their API, which now accepts 100 images per request, up from 8 for previous Mistral models Key model details: ➤ Context Window: 512k tokens ➤ Multimodality: Text and image input, with text output ➤ Pricing: $1.36/$4.18 per 1M input/output tokens ($0.14 per 1M cached input tokens), with 50% off for the first two weeks ($0.68/$2.09) ➤ Availability: Research Public Preview on Mistral's API, with open weights planned for the end of October
#模型发布#评测/基准#开源生态
Anthropic 发布 Claude Haiku 5.5，这是首个带 effort 设置与自适应思考的 Haiku 模型，并引入分层定价：100k token 以内为 $0.10/$0.50 每 1M 输入/输出 token，超过 100k 后升至 $0.50/$2.50。Artificial Analysis 评测显示，Haiku 5.5（max）在小型模型中略领先 GLM-5.3 Flash（42）、Gemini 3.8 Flash（41）和 GPT-6 Luna（38），但落后 Claude Sonnet 5.5（max，56）13 分；其上下文窗口从 Claude 4.5 Haiku 的 200k 提升到 1M token。该模型每个 Intelligence Index 任务约消耗 162k 输出 token，约为 GPT-6 Luna（max，约 50k）的 3 倍，Terminal-Bench 4.0 得分 33%（Haiku 4.5 为 0%）。
#模型发布#Anthropic#评测/基准