最新模型

新模型、開放權重、多模態、語音、影像與模型架構。

NVIDIA Dynamo Prewarms a Standby Engine with Shared HBM Weights, Cutting Failover from 283 Seconds to 7.3 Seconds

Dynamo’s experimental Shadow Engine Failover lets two inference processes map the same GPU weights, avoiding a model reload after a process crashes. NVIDIA’s GLM-5.2 test showed a nearly 39× reduction in recovery time, but the feature currently primarily supports vLLM and cannot handle GPU, node, or cross-node failures.

2026-08-26

最新模型

Tencent Open-Sources WeMM-Embedding: 2B Model Outperforms Some 8B Baselines in Multimodal Retrieval Benchmarks

WeMM-Embedding processes text, images, video, visual documents, and interleaved inputs in a shared vector space and is available in 2B, 4B, and 9B sizes. Official results show it outperforming several public baselines, but the model does not support audio, and some product-level results are backed only by Tencent’s internal A/B tests.

2026-08-26

訓練資料

LAION-BVD Releases a 10-Million-Hour Video Corpus, but Media Files Are Limited to Approved Noncommercial Research

LAION-BVD compiled 1.3 billion video URLs from Common Crawl and downloaded approximately 80 million videos totaling 10 million hours. The data pipeline and URL lists are public, but access to the main media subset is restricted, and researchers remain responsible for compliance.

2026-08-26

AI 安全

JFrog Routes Coding Agent Package Installs Through Artifactory, but Routing Is Disabled by Default in the Preview

Agent Package Resolution uses session prompts, package manager configuration, and server-side policies to direct dependency downloads triggered by Claude Code and Cursor to Artifactory. The feature remains in preview, with no repository bindings in its initial configuration, and the plugin alone cannot prevent agents from bypassing it to access public registries.

2026-08-26

代理工具

Recuris Freezes Agent Weights and Recursively Rewrites Skill Memory, Improving Results in 35 Long-Horizon Task Evaluations

Recuris separates task state, skill selection, and failure evidence into memory components that can be updated independently, then validates each modification against held-out data. The authors recorded improvements in 35 of 37 completed model–benchmark pairings, although some gains were not statistically significant and the evaluations also relied on LLM-simulated users and judges.

2026-08-26

AI 開發平台與治理

OpenAI Admin Plugin Turns Usage, Permissions, and Member Management into Agent Tools—Without Creating a New Authorization Boundary

OpenAI has introduced the Admin plugin for ChatGPT Work and Codex, allowing administrators to query usage conversationally and perform supported management operations. The plugin inherits existing roles, App permissions, and source-system permissions; installation alone does not grant access to data or write privileges.

2026-08-26

最新模型

EchoWM Integrates Six-Degree-of-Freedom Navigation and Native Audio Into a World Model, but the Interactive Code Has Yet to Be Released

The JD.com Joy Future Academy team has introduced EchoWM, which uses a unified trajectory interface to control 720p audiovisual worlds from both first- and third-person perspectives. The paper describes the system as an “open” model, but the currently linked repository still provides only a long-video pipeline; EchoWM’s code and weights have yet to be released.

2026-08-26

最新模型

IBM Granite 4.2 Opens Three Dense Reasoning Models, but Official Documents Disagree on Their Training Starting Point

Granite 4.2 introduces 3B, 8B, and 30B models with switchable thinking modes, native tool calling, and context windows of up to 512K, under the Apache 2.0 license. The 8B and 30B models underwent sandbox-based agentic reinforcement learning, but the model card and technical article contradict each other on whether the models were pretrained from scratch.

2026-08-26

代理工具

Pipette Open-Sources On-Device Model Benchmarking for Accuracy, Latency, and Memory—but Rankings Still Cannot Be Compared Directly Across Devices

Liquid AI and Artificial Analysis have released Pipette, which measures on-device inference across more than 1,000 combinations of models, quantizations, runtimes, devices, and context configurations. Speed and memory are measured on physical devices, while quality scores are generated separately on H100 GPUs, so the two testing paths must be distinguished when interpreting its combined charts.

2026-08-26

AI 安全

TRACE Joins the Linux Foundation: Using Hardware Attestation to Prove Where an Agent Runs, but Not That Its Behavior Is Correct

The Linux Foundation has accepted the TRACE specification, which creates portable cryptographic evidence for AI agents’ models, execution environments, policies, and tool calls. The current v0.2 remains a Developer Preview: it can prove that records are linked to an attested environment, but not that an agent completed its task or that the policy itself is secure.

2026-08-26

代理工具

TRACE Rewrites Agent Skills from Successful and Failed Trajectories, Raising the Consistent Pass Rate on Hidden Tests to 70%

Xiaomi’s team introduced TRACE, which iteratively updates a retrievable Skill Bank by contrasting successful and failed trajectories without modifying the underlying model weights. It increased the share of CAR-bench hidden-set tasks passed in all three runs, but also raised average token usage, cost, and latency.

2026-08-25

最新模型

NVIDIA Adds Two-Stage Generation for MiniMax H3, Cutting 10-Second Video Latency to 14.9 Seconds on a Single GB200

H3 Super Acceleration first creates a low-resolution draft with four denoising steps, then uses LTX and Sol-Attn to complete high-resolution refinement in three steps. NVIDIA reports that 768p, 10-second video generation is 27.7× faster than the SGLang baseline, but the acceleration changes the sampling path and is not a pure kernel optimization at equivalent quality.

2026-08-25