搜索“Agent”

ore tokens than GPT-6 Luna: Haiku 5.5 (high) scores 38 with 55k tokens per task against 38 with 50k for Luna (max), and the gap widens at lower effort settings ➤ Highly capable at agentic knowledge work: on AA-Briefcase, our private evaluation for realistic knowledge work tasks, Haiku 5.5 (max) reaches 1578 Elo, ahead of models including Kimi K3 and GLM-5.3, and comparable to Muse Spark 1.3 (max) ➤ Improvements on terminal use: on Terminal-Bench 4.0 it scores 33%, up from 0% for Haiku 4.5. This

匹配字段正文

score near zero on CyberGym-E2E because they refuse outright. Reproducing a vulnerability to prove it is real is standard defensive work, and provider-level refusals block it. In agentic coding, Mistral reports 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA, and 28.3% on Terminal-Bench 4.0, for a combined Artificial Analysis Coding Agent Index of 49.8%. Mistral notes these were evaluated privately ahead of the harness going public, so they are not yet independently reproducible. A blind human ev

匹配字段正文

ingGemma 2 need? Google reports about 191MB of active RAM for text-only use and 567MB for full multimodal use, quantized, on a Pixel 11 Pro [Sponsored] The web is the one API most agents are missing. Databases, calendars and repos have APIs. The open web mostly doesn’t. The TinyFish MCP server gives any MCP client four tools: TinySearch, TinyFetch (full pages as markdown, JavaScript included), TinyBrowser for logins and forms, and TinyAgent for multi-step jobs. Search and Fetch are free. Asif Ra

匹配字段正文

g a significant 9.92-point improvement on code performance (in MTEB Code, from 68.76 to 78.68), making it well-suited for local codebase indexing, semantic code search, and coding agent retrieval. Across image, video, documents, and audio, it sets a new standard in quality-per-parameter for sub-1B models and even outperforms some specialist models more than twice its size. Find full evaluation metrics and model information in the EmbeddingGemma 2 model card. Enabling semantic search, routing, an

匹配字段正文

g a significant 9.92-point improvement on code performance (in MTEB Code, from 68.76 to 78.68), making it well-suited for local codebase indexing, semantic code search, and coding agent retrieval. Across image, video, documents, and audio, it sets a new standard in quality-per-parameter for sub-1B models and even outperforms some specialist models more than twice its size. Find full evaluation metrics and model information in the EmbeddingGemma 2 model card. Enabling semantic search, routing, an

匹配字段正文

4 is designed to run in private clouds or on-premise as well. At the same time, Mistral touts high refusal rates. On malicious cyber prompts from JailbreakBench, StrongREJECT, and AgentHarm, ML4 refuses more often than any other open model. How the model reliably tells legitimate vulnerability research apart from attack prep, Mistral doesn't explain. In Lakera's B3 AI Security Benchmark, ML4 blocks 93.3 percent of attacks, according to the company. ML4 takes second place in coding, ahead of Deep

匹配字段正文

ve parameters. It is our largest and most capable model to date, and it continues to improve rapidly as we refine it. The model demonstrates exceptional performance across coding, agentic workflows, and multimodal understanding. It already achieves performance competitive with the strongest open-source models globally, while significantly outperforming any open-weight model developed in the US or Europe. On critical enterprise workloads, including cybersecurity, finance and law, we find it to be

匹配字段正文

ong results with minimal compute, putting it in direct competition with Chinese open-weight models from Deepseek and Qwen. Reflection built Beam for coding, logical reasoning, and agentic tasks. The mixture-of-experts model activates 23 billion of its 501 billion total parameters per token, keeping compute costs relatively low. On demanding reasoning tasks, Beam matches GLM 5.2 while using three to four times less compute, according to Reflection. The company is going after businesses that use A

匹配字段正文

del race Image source: Reflection The Rundown: U.S.-based Reflection AI, the $25B startup by former DeepMind researchers, just introduced Beam, an open-weight model for coding and agents that the company says competes with Chinese rivals at lower cost and greater efficiency. The details: Reflection claims Beam needs a fraction of the computing power of Chinese lab Z.ai’s GLM-5.2, but achieves similar scores across reasoning, coding, and more; Moonshot’s Kimi K3 still leads Beam on raw ability, b

匹配字段正文

Ling 3.1 Flash makes large gains in intelligence over its predecessor, scoring 41 on the Artificial Analysis Intelligence Index with particular improvement in agentic capabilities @AntLingAGI has released Ling 3.1 Flash, a reasoning model with 560B total parameters, 25B active parameters, and a 1M token context window. It scores 41 on the Artificial Analysis Intelligence Index v4.3, up from 20 from Ling 3.0 Flash, which launched in August. Ling 3.1 Flash is expected to be open weights, with Ant

匹配字段正文

ed from scratch. A single neural network understands and generates text, images and video, reasons over them, and outputs robot actions. Reka frames it as a direct replacement for agentic pipelines that pass work between modality-specific models. What Rho-1 Changes Most multimodal systems today are pipelines. A central model plans, then hands jobs to specialists for images, video or detection. Each handoff adds latency, and each specialist sees only a narrow request. Rho-1 removes those handoffs

匹配字段正文

nder, AINews is now a section of Latent Space. You can opt in/out of email frequencies! Beam launch: Reflection announced Beam, a text-only 501B-total / 23B-active MoE for coding, agentic and scientific work. It was trained from scratch, and full weights under Apache 2.0 are due this month (announcement, Laskin).Training scale: Team posts cite 23.8T pretraining tokens, partly from an OCR pipeline over hundreds of millions of PDFs (data lead). They also describe a stable RL/OPD run on 10K GB300s

匹配字段正文

ional vector space; Superior code understanding: Significantly outperforms EmbeddingGemma 1 on code search and technical retrieval, making it ideal for local codebase indexing and agentic code search; Modular memory footprint: Selectively load only the encoders you need at runtime: 270M (text/code), 440M (text + vision), 570M (text + audio), or 740M (full multimodal), all projecting into the same compatible vector space; Flexible vector storage: Matryoshka Representation Learning (MRL) enables d

匹配字段正文

l Dua, Sandeep Patil, Sebastian Russo, Sebastian Schmidt, Shengyi Lin, Shuangfeng Li, Somdatta Banerjee, Steven Toribio, Suril Shah, Sylvain Vignaud, Tenghui Zhu, Vinod Mamillapalli, Vladimir Kirilyuk, Wai Hon Law, Weiyi Wang, Xiaoming Hu, Xinan Cheng, Xu Chen, Yi-Chun Kuo, Yishuang Pang, Yu-hui Chen. Optimizing a 200,000-line Python codebase overnight with Gemini 4 Argon Gemini 4 Argon for Robotics Announcing ADK for Kotlin 1.0: Building Production-Ready AI Agents in Kotlin, Android, and Beyond

匹配字段正文

introduced Beam, its first open-weight model. Beam is a sparse Mixture-of-Experts (MoE) model with 501B total parameters and 23B active per token, built for coding, reasoning and agentic workloads. As per the Reflection AI team, Beam directly competes with larger open models like GLM 5.2 while using 3 to 4x less inference compute on reasoning benchmarks. Is it deployable today? Not for self-hosting yet. Beam is in final red-teaming. Early access runs through a waitlist on the Reflection platfor

匹配字段正文

a lengthy blog post Monday , which described Beam as a text-only mixture-of-experts model trained on high-compute reinforcement learning to be effective at reasoning, coding, and agentic tasks at “a fraction of the token cost and inference time compute” of rivals. Beam is a 501-billion-parameter model with 23 billion active parameters. It was pretrained on 23.8 trillion tokens and has a 1 million token context window. To compare, Z.ai’s GLM-5.2 has roughly 744 billion total parameters with 40 b

匹配字段正文

% of cases. Accounting and numerical precision bugs add 18%. Time validation, business rules and SSRF cover the rest. As per the model card on HF, RL rollouts ran inside the Codex agent harness on production-like software and protocol environments. Benchmark Results Cantina evaluated 60 tasks from 20 held-out vulnerability cases. Each model ran the set once, with costs estimated from provider pricing. apex-flash-1: 40/60 solved (66.7% pass@1), about $2.38; GLM-5.3-Flash (base): 36/60 solved (60.

匹配字段正文

s a question; Smart Folders: d1 files each new document into a folder, then a subfolder. It files search questions the same way, so keyword search only looks in one subfolder; Web Agent: d1 operates a flight-search website from a one-sentence goal. At each step, it picks the next action among everything the page allows; Context Compaction: d1 reads each tool output in a coding agent's session and keeps, trims, or drops it for the next task. It removes 52% of the tokens and keeps every output the

匹配字段正文

luated every model with the same eval-framework setup. In English, Kolibri leads GPQA Diamond (84.3), AIME 2025 (96.9) and AIME 2026 (96.0). It ties Qwen3.5 35B-A3B on the English agentic average at 63.4. It trails on BFCL v4, scoring 61.4 against 70.5 for Qwen3.5. The dense Qwen3.8 27B scores higher overall (80.2 EN, 79.9 DE), but activates about 8 times more parameters per token. Against its internal predecessor Kolibri Origin, Kolibri decodes about 2.7 times more text per GPU while scoring 21

匹配字段正文

ce actually supports rather than quietly broadening a study’s conclusions, and researchers need to be able to verify the final outputs. We see that in how scientists use Asta, our agentic platform for scientific work. Instead of simple keyword searches, users often bring substantial context and many constraints—for example, asking Asta to compare approaches across a body of literature while accounting for a particular method, population, or setting. Many also return to generated reports later, t

匹配字段正文