搜索“Agent”

We chose to evaluate SWE-1.6 on SWE-Bench Pro following OpenAI’s recommendation as the spiritual successor of SWE-Bench Verified. Running bug-free and reproducible evaluations for agents requires care. We manually read hundreds of trajectories and cross-checked against Scale-reported SWE-Bench Pro trajectories when applicable. Some examples of subtle issues we resolved include: dependency issues in agent and grading environment setup, inconsistent handling of timeouts across harnesses, edge case

匹配字段正文

Claude Sonnet 4.6 is our most capable Sonnet model yet. It’s a full upgrade of the model’s skills across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. Sonnet 4.6 also features a 1M token context window in beta. For those on our Free and Pro plans, Claude Sonnet 4.6 is now the default model in claude.ai and Claude Cowork. Pricing remains the same as Sonnet 4.5, starting at $3/$15 per million tokens. Sonnet 4.6 brings much-improved coding skills to more

匹配字段正文

略优化，模型能够以高吞吐处理长达 1M token 的上下文。; 更高的 token 效率：引入“正确性 + 过程冗余”复合奖励机制，进一步拓展了即时模型效率与效果的平衡边界。在相同 token 效率条件下，Ling-2.5-1T 的推理能力显著超越前代，接近需消耗约 4 倍输出 token 的前沿思考模型水平。; 精细化偏好对齐：通过引入双向强化学习反馈、Agent-based 指令约束校验等精细化对齐策略，使 Ling-2.5-1T 在创意写作、指令遵循等偏好对齐类任务上相比前代模型实现大幅提升。; 高效的原生智能体交互：基于大规模高保真交互环境进行 Agentic RL 训练，Ling-2.5-1T 可适配 Claude Code、OpenCode、OpenClaw 等主流智能体产品。在通用工具调用基准 BFCL-V4 上达到开源领先水平。 我们全面评估了 Ling-2.5-1T 在知识、推理、智能体交互、指令遵循、长文本处理等多个权威基准评测上的表现。Ling-2.5-1T 与前代模型 Ling-1T 对比，实现了全方位的能力提升，是百灵家族当前最强大的即时模型。在与主流的大尺

匹配字段正文

Qwen3.5: Towards Native Multimodal Agents | Qwen Blog Publication About Try Qwen Chat Qwen3.5: Towards Native Multimodal Agents February 14, 2026 · 196 min · 41706 words · Qwen Team | Translations: 简体中文 QWEN CHAT GitHub Hugging Face ModelScope DISCORD We are delighted to announce the official release of Qwen3.5 , introducing the open-weight of the first model in the Qwen3.5 series, namely Qwen3.5-397B-A17B . As a native vision-language model, Qwen3.5-397B-A17B demonstrates outstanding results ac

匹配字段正文

制，在超过 32K 生成长度下，访存规模降低 10 倍以上，生成吞吐提升 3 倍以上，尤其适合深度思考和长程执行的任务。; 深度思考：在 RLVR 基础上引入 dense reward 来反馈思考过程的严谨性，使得 Ring-2.5-1T 同时实现 IMO 2025 和 CMO 2025 的金牌水平（自测）。; 长程执行：通过大规模 fully-async agentic RL 训练，显著提升针对复杂任务的长程自主执行能力，使得 Ring-2.5-1T 可以轻松适配 Claude Code 等智能体编程框架和 OpenClaw 个人 AI 助理。 1. 深度思考与长程执行 为评估 Ring-2.5-1T 的深度思考和长程执行能力，我们选取了具有代表性的开源思考模型（DeepSeek-v3.2-Thinking、Kimi-K2.5-Thinking）和闭源API（GPT-5.2-thinking-high、Gemini-3.0-Pro-preview-thinking-high、Claude-Opus-4.5-Extended-Thinking）作为参考。 Ring-2.5-1T 在数学

匹配字段正文

re introducing our latest model, MiniMax-M2.5. Extensively trained with reinforcement learning in hundreds of thousands of complex real-world environments, M2.5 is SOTA in coding, agentic tool use and search, office work, and a range of other economically valuable tasks, boasting scores of 80.2% in SWE-Bench Verified, 51.3% in Multi-SWE-Bench, and 76.3% in BrowseComp (with context management). Trained to reason efficiently and decompose tasks optimally, M2.5 exhibits tremendous speed in performi

匹配字段正文

ned to deliver natural, expressive and production-ready voices for Indian languages. Voice is how India uses technology. Digital-native platforms onboard gig workers through voice agents, with no forms and no apps to download, just conversation. Students ask AI tutors to explain concepts in their native language. Banks handle customer queries through voice at volumes that would be impossible for human agents alone. Gamers interact with characters who speak back in their language. These experienc

匹配字段正文

tsIntroducing Claude Opus 4.6Feb 5, 2026 We’re upgrading our smartest model. The new Claude Opus 4.6 improves on its predecessor’s coding skills. It plans more carefully, sustains agentic tasks for longer, can operate more reliably in larger codebases, and has better code review and debugging skills to catch its own mistakes. And, in a first for our Opus-class models, Opus 4.6 features a 1M token context window in beta1. Opus 4.6 can also apply its improved abilities to a range of everyday work

匹配字段正文

We’re upgrading our smartest model. The new Claude Opus 4.6 improves on its predecessor’s coding skills. It plans more carefully, sustains agentic tasks for longer, can operate more reliably in larger codebases, and has better code review and debugging skills to catch its own mistakes. And, in a first for our Opus-class models, Opus 4.6 features a 1M token context window in beta1. Opus 4.6 can also apply its improved abilities to a range of everyday work tasks: running financial analyses, doing

匹配字段正文

ion, context biasing, and word-level timestamps in 13 languages; Voxtral Realtime: Purpose-built for live transcription with latency configurable down to sub-200ms, enabling voice agents and real-time applications; Best-in-class efficiency: Industry-leading accuracy at a fraction of the cost, with Voxtral Mini Transcribe V2 achieving the lowest word error rate, at the lowest price point; Open weights: Voxtral Realtime ships under Apache 2.0, deployable on edge for privacy-first applications Voxt

匹配字段正文

ieves significant performance improvements across multiple dimensions, including factual knowledge, complex reasoning, instruction following, alignment with human preferences, and agent capabilities. On 19 established benchmarks, it demonstrates performance comparable to leading models such as GPT-5.2-Thinking, Claude-Opus-4.5, and Gemini 3 Pro. We further enhance Qwen3-Max-Thinking with two key innovations: (1) adaptive tool-use capabilities that enable on-demand retrieval and code interpreter

匹配字段正文

.5-1.2B-Thinking outpaces not only Qwen3-1.7B but even hybrid architectures like Granite-4.0-H-1B in both speed and memory efficiency. We found that LFM2.5-1.2B-Thinking shines on agentic and reasoning-heavy tasks (e.g., tool use, math, programming). When the model needs to plan a sequence of tool calls, verify intermediate results, and adjust its approach, the reasoning traces provide real value. However, we recommend using LFM2.5-1.2B-Instruct for chat capabilities and creative writing. Traini

匹配字段正文

ages, as follows. (1) Supervised Fine-Tuning (SFT): Learning to understand and responding to a variety of instructions As we transition from the era of generative AI to the era of agentic AI, teaching AI to use various tools has become a critical challenge. To achieve this, environments where AI can directly connect and interact with real-world tools and APIs must be created, a process that is costly and time-consuming. Therefore, K-EXAONE created a virtual tool usage environment instead of real

匹配字段正文

del family, our most capable release yet for edge AI deployment. It builds on the LFM2 device-optimized architecture and represents a significant leap forward in building reliable agents on the edge. We've extended pretraining from 10T to 28T tokens and significantly scaled up our post-training pipeline with reinforcement learning, pushing the boundaries of what 1B models can achieve. This is a comprehensive release with Base, Instruct, Japanese, Vision-Language, and Audio-Language models. LFM2.

匹配字段正文

ral OCR 3 is ideal for both high-volume enterprise pipelines and interactive document workflows. Developers can use it for: Extracting text and images into markdown for downstream agents and knowledge systems Automated parsing of forms, invoices, and operational documents End-to-end document understanding pipelines Digitization of handwritten or historical documents Any other document → knowledge transformation applications. Our early customers are using Mistral OCR 3 to process invoices into st

匹配字段正文

, and complex logic across 40+ native languages. Scalable efficiency: From 3B to 675B parameters, choose the model that fits your needs, from edge devices to enterprise workflows. Agentic and adaptable: Deploy for coding, creative collaboration, document analysis, or tool-use workflows with precision. Explore the model documentation: Technical documentation for customers is available on our AI Governance Hub Start building: Ministral 3 and Large 3 on Hugging Face, or deploy via Mistral AI’s plat

匹配字段正文

th supervised fine-tuning and large-scale reinforcement learning of Mixture-of-Experts models; Verifiers and the Environments Hub: A unified environment interface and ecosystem foragentic RL training environments and evaluations; Prime Sandboxes: High-throughput, secure code execution for agentic coding environments; Compute Orchestration: Orchestrated and managed 512 NVIDIA H200 GPUs across 64 interconnected nodes PRIME-RL INTELLECT-3 was trained end-to-end with prime-rl, our production-scale p

匹配字段正文

Our newest model, Claude Opus 4.5, is available today. It’s intelligent, efficient, and the best model in the world for coding, agents, and computer use. It’s also meaningfully better at everyday tasks like deep research and working with slides and spreadsheets. Opus 4.5 is a step forward in what AI systems can do, and a preview of larger changes to how work gets done. Claude Opus 4.5 is state-of-the-art on tests of real-world software engineering: Opus 4.5 is available today on our apps, our AP

匹配字段正文

h the help of our close partners, we look forward to substantially scaling up our training efforts and training AI models that will deliver the next generation of breakthroughs in agentic capabilities, long term memory, and continual learning. In our accompanying technical report, we provide a deep dive into our experience and insights from our efforts pretraining on AMD hardware as well as describe our training framework and the hardware specific optimizations we perform in detail. We aim to di

匹配字段正文

TL;DR: We trained SWE-grep and SWE-grep-mini, fast agentic models specialized in highly parallel context retrieval. They match the retrieval capabilities of frontier coding models, while taking an order of magnitude less time. Available now in Windsurf’s new Fast Context subagent! Modern coding agents face a fundamental tradeoff between speed and intelligence. Frontier models can solve complex tasks, but it can take minutes of searching before they edit a single file, breaking your flow. In Wind

匹配字段正文