#Agent

AIOpenBMB released MiniCPM5-2B-SFT, an SFT-only BF16 checkpoint taken before RL and OPD, within its MiniCPM5-2B series. The model is a 2B dense Transformer built for on-device and local deployment, with 131,072-token context and the same training recipe as the final release.

Why it matters: The source gives concrete benchmark averages against same-size and larger models, plus released training data and multiple deployment formats, useful for judging a compact on-device model.

AIOpenBMB released MiniCPM5-2B, a dense 2B Transformer for on-device and resource-constrained deployment, alongside its training datasets. The source reports a 53.9 average across its comparison set and strong results in coding, math, long-context, tool use, and agentic tasks. This page is the pre-training base checkpoint, with BF16 weights and GGUF, MLX, GPTQ, and LiteRT-LM variants listed separately.

AIQwen has published Qwen-Drive-1.0-4B on Hugging Face, a vision-language model for autonomous driving built on Qwen3.5-4B. The release includes a BEV perception head and two Planning Experts, planner-sft and planner-rl, with code and an inference example in the linked GitHub repository.

Why it matters: The source gives concrete benchmark results and a runnable setup, letting readers judge how a driving VLM with planning and perception heads compares with existing systems.

AITencent has published ContextPilot-E4B on Hugging Face, the Gemma4-E4B checkpoint of ContextPilot, a framework that teaches long-horizon language-model agents to plan, maintain long-term memory, and offload less useful context while reasoning and using tools. The checkpoint is intended for research on proactive context management, long-context QA, and deep search, and loading it alone does not execute the context-management tools, which are provided in the ContextPilot repository.

AITencent has released ContextPilot-14B on Hugging Face, a Qwen3-14B checkpoint for proactive context management in long-horizon language-model agents. The framework lets agents plan, maintain long-term memory, and offload less useful context while reasoning and using tools. The checkpoint is intended for research on long-context QA and deep search, and loading it alone does not execute the context-management tools, which are provided in the ContextPilot repository.

AIIn an interview with Jiazi Guangnian, Renmin University information school dean Chai Yunpeng describes his team's social simulator, which runs over 13.5 million AI agents calibrated against the CGSS survey data. He argues that social world models are the missing piece for AI agents that must interact with people, and that the startup Jingtong Technology has raised two funding rounds in two months.

AINVIDIA and AWS are expanding their partnership across GPUs, CPUs, networking, open models and software. The announcement cites 2 million additional NVIDIA GPUs across AWS infrastructure, the NVIDIA Vera CPU coming to AWS for agentic AI, NVLink Fusion with NVHBM memory, and 100,000 GPUs for U.S. government AI factories on secure AWS infrastructure.

AI…databricks! Agentic development will need a new kind of cloud infra.

AI…technologies they developed for communication and coordination, and interactive figures analyzing agent activity:

AIThe most common was to learn how the ExploitGym scorer works in order to trick or tamper with it. Other rationales included finding specific task solutions and obtaining shared infrastructure or credentials.

AIWe ran a sweep looking for this, and a representative example of the most severe cases we found was an agent writing a malicious pull request with a misleading description.

AIPHASEONE10841 passed along its work to PHASEONE[big], which had the same task but a larger budget. PHASEONE[big] was “poisoned” so it was particularly interested in tampering with logs.

AIThey recruited “sacrificial” agents to deliberately end their run and submit to trigger the tripwire and generate information for the “collective”.

AIThey successfully developed a method to pretend to run one command while actually running another (“tool call spoofing”).

AIGrok Bot is now available to all standard Grok and Cursor subscribers, with SuperGrok and Cursor Pro subscribers included. Cursor's Michael Truell says users are delegating tasks ranging from running small e-commerce businesses to testing production software. Weekly usage limits are also being reset for all users.

AIWatch how this latest model filters out spoken hesitation while grounding technical terms, file names, and code variables precisely against the active context while building in @Antigravity. By using visual biasing, the model brings deep, screen-aware context to complex developer workflows.

AIZ.ai released GLM-5.3-Flash, a 320B-A18B model, with day-0 support in SGLang, after appearing earlier as ox-alpha. The post calls it the first native multimodal model in the GLM-5 series and says it outperforms GLM-5.2 at one-tenth the cost, with stable 1M-token long-context performance.

Why it matters: The post reports GLM-5.3-Flash's native multimodal design, its efficiency claims, and day-0 SGLang support, which bear on running it in practice.

AIDeepSeek V4 Pro 0813 scored 95.2% on SWE-Bench Verified, ahead of Kimi K3 at 92.6% and Fable 5 at 85.4%, in Fireworks AI's eval runs. It costs $0.309 per solved task on SWE-bench versus $0.808 for Fable 5, and it is available through Fireworks serverless and dedicated endpoints, with SFT, DPO, and RFT training support. Its 1M-token context window and native tool calling target long-horizon agentic workloads, though its Java accuracy on Aider Polyglot (48.9%) trails Fable 5 (74.5%).

AIDeepSeek V4 Pro 0813 recorded zero refusals across 840 adversarial security tasks in CyberGym testing, solving them at about half the cost per success of the top-scoring model tested, Kimi K3. In the 697-task common cohort, V4 Pro reached a 53.7% reward rate at $2.50 per solved task, versus 47.6% and $9.64 for GPT-5.5 and 5.9% and $33.28 for Claude Opus 4.8.

AIHarvey and Fireworks post-trained Tenet from the Kimi K3 base using asynchronous reinforcement learning on the Fireworks Training API for long-horizon legal work. On the Legal Agent Benchmark, Tenet reached 19.7% all-pass versus 10.8% for base Kimi K3, and its cost per task was $5.92 versus $5.62.

AIZ.ai has released GLM-5.3-Flash, a model with native visual capabilities that observe interfaces, rendering results, and interaction feedback across code, browsers, and GUIs. It uses a hybrid linear and sparse attention architecture with 320B total parameters and 18B activated, which the company says significantly reduces compute and KV-cache requirements. The release notes also describe support for office document and financial research workflows.

Why it matters: The release notes give GLM-5.3-Flash's architecture, parameter counts, and cybersecurity findings, which make the model's scope concrete for comparison with earlier GLM releases.

AI…approval, then resume • Runs anywhere Node.js runs

AIOpenWorker, an open source agent that completes tasks on a laptop, has released a new version with built-in cybersecurity agents. The agents scan code for vulnerabilities, scan dependencies for supply chain injections, and check cloud security configurations for attack surfaces. Users can run open weight models locally so sensitive code stays on their machine.

AIVercel Connect is now generally available, letting apps and agents securely access more than 100 services, including Slack, Linear, and GitHub. It uses short-lived, scoped access tokens, token and trigger observability, and RBAC with audit trails.

AIZ.ai released GLM-5.3 on Hugging Face, built on the same base model as GLM-5.2, with all gains coming from post-training. The source reports a 50% improvement over GLM-5.2 on Z.ai Code Bench and open-source SOTA on Terminal Bench 3.0 and Agents' Last Exam, with a benchmark table comparing it against Kimi K3, DeepSeek-V4 Pro-0813, Qwen3.8-Max, and others.

Why it matters: The source gives benchmark tables against GLM-5.2 and rival models, showing where the post-training gains concentrate in coding and cyber tasks.

AIGrok diagnoses hardware issues, ships replacements, and fulfills over 3,000 orders a week across voice calls and chat.

AIMicrosoft's AI Blog outlines five signals that trust, not speed alone, lets organizations scale AI from pilots to enterprise-wide use. Its first signal is observability, citing Microsoft's Cyber Pulse AI Security Report finding that 29% of employees use unsanctioned AI agents their security teams cannot see. The post also says security should be built into AI systems by design and governance should be continuous rather than a one-time approval.

AIit's been inspiring to see science and engineering get better faster as our team enables collaboration between agents and people 🦾

AISwyx argues that simulating humans and their feedback is likely the final barrier to recursive self-improvement, where models automate increasingly large parts of ML research. He says Simile, which builds human simulations, is already finding product-market fit with Fortune 100 companies despite its early stage.

AITo help us understand its real-world behavior, we are making it available for free on OpenRouter (only with agentic harnesses) for the next few weeks, starting now. We’ll use the data, disassociated from accounts, to better it.

AIAndrew Ng's AI Engineering Skills Map ranks building and deploying AI applications as the top skill tier, spanning LLM foundations, data grounding, agentic systems, evaluation-driven development, production operations, and machine learning foundations. He explains that AI outputs are less predictable than traditional software, so skilled engineers build iteratively, examining results and deciding next steps based on intermediate outcomes. The skills map was derived from job postings, expert interviews, and survey responses.

AIAmazon Science released SOP-Bench, an open benchmark that measures how well AI agents execute standard operating procedures written by domain experts. It covers 12 business areas, including healthcare intake and dangerous-goods classification, with more than 2,000 tasks, working tools, and ground-truth answers. The benchmark was presented at the 2026 KDD conference.

AI…stack) and now in the AI Engineering era seeing SF builders rediscover MDB was so gratifying - this was a *huge* step up from last year!

AIV4-Flash-Vision-Exp works smoothly across agent frameworks, combining visual understanding with a wide range of tools to unlock more practical workflows. 2/n

AIDeepSeek has made its experimental multimodal model DeepSeek-V4-Flash-Vision-Exp available on the DeepSeek API Platform. The company says it matches DeepSeek-V4-Flash on text tasks, including agents, reasoning, and world knowledge. On multimodal agent benchmarks it improves substantially over V4-Flash and approaches Opus-4.8, and DeepSeek Harness 0.1.1 was released the same day with support for the new model.

AIDeepSeek has made DeepSeek-V4-Flash-Vision-Exp, an experimental multimodal vision understanding model, available on its API platform via model='deepseek-v4-flash-vision-exp'. The source says its pure-text capabilities are on par with DeepSeek-V4-Flash, while it shows a significant leap on agent benchmarks requiring visual understanding, which it says brings multimodal agent capabilities close to Opus-4.8.

Why it matters: The source gives benchmark scores and a model identifier, so readers can compare the experimental vision model against the text-only DeepSeek-V4-Flash on agent tasks.

AIThe author argues that agents editing their own harnesses and training loops are only bounded self-improvement so far. Recursion would require the system to also raise and keep an honest evaluation bar, which current evidence does not show.