AI 每日快讯

AI 每日快讯

AI 产品、模型、开源工具和官方动态的时间流。保留历史记录,按分类、日期和标签继续筛选。

4975历史快讯
269开源工具
80当前结果
11 月 16 日 昨日快讯
MarkTechPost 官方资讯

MarkTechPost:Meet Meta’s JiTTests: LLM-Generated Tests Designed to Fail Before Risky Code Lands

原文摘要:Meta has moved Catching JiTTests into production. The LLM system writes tests designed to fail on risky diffs, and caught 74 privacy SEVs before landing in H1 2026, with a 60% engi 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 15 日 2026-11-15 快讯
11 月 14 日 2026-11-14 快讯
MarkTechPost 官方资讯

MarkTechPost:Stanford Research Open-Sources Memorilla, a Latent Memory Module for Frozen LLMs

原文摘要:Stanford researchers released Memorilla, an open-source memory system that compresses a document collection into 16 memory tokens for a frozen Qwen3-8B. It reaches 84.33% on FactKG 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 13 日 2026-11-13 快讯
MarkTechPost 官方资讯

MarkTechPost:Best AI Video Generation Models in 2026: Elo, Price per Second, Open Weights

原文摘要:The AI video leaderboard looks very different from 12 months ago. Native audio is now table stakes. Clips run 15–30 seconds. And an open-weights model sits in the top 3. This ranki 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 12 日 2026-11-12 快讯
MarkTechPost 官方资讯

MarkTechPost:CopilotKit Releases OpenBot 1.0: An MIT-Licensed, Self-Hostable AI Coworker With Built-In Go…

原文摘要:CopilotKit released OpenBot 1.0, an MIT-licensed, self-hostable AI coworker workspace. Companies bring their own agents, models and infrastructure, while every Bot action passes a 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:AWS Releases AWS Emulator: A Local Copy of 200+ AWS Services Built by AI Agents

原文摘要:AWS released AWS Emulator, a free, CC0-licensed local emulator of AWS APIs built from 438 API models. Autonomous agents wrote all of its code. It passes 56.6% of 42,857 published f 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 11 日 2026-11-11 快讯
11 月 10 日 2026-11-10 快讯
MarkTechPost 官方资讯

MarkTechPost:Meta FAIR Releases e-means: An Algorithm Beating k-means in 10x Fewer Epochs

原文摘要:Meta FAIR’s e-means is a mini-batch k-means replacement shipped in FAISS 1.16.0. It reaches Lloyd’s 25-epoch loss in 1.4 to 2.6 epochs, with 10-45x GPU speedups and no extra hyperp 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Anthropic Launches Claude Fable 5.2 and Mythos 5.2: Same Model, Different Safeguards

原文摘要:Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, one model with two safeguard levels. Fable 5.2 scores 49.1% on FrontierCode 2.0, costs $8/$40 per million tokens, and shi 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 09 日 2026-11-09 快讯
MarkTechPost 官方资讯

MarkTechPost:Baidu’s ERNIE Evolve Automated Research System Beats NanoChat and NanoGPT Records

原文摘要:Baidu’s ERNIE Evolve is an automated research system that edits code, runs experiments and keeps only verified gains. It beat public bests on NanoChat and NanoGPT Speedrun and sped 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 08 日 2026-11-08 快讯
MarkTechPost 官方资讯

MarkTechPost:Which Routing Does Each MoE Model Use? DeepSeek, Qwen, Mixtral, Llama 4, gpt-oss Compared

原文摘要:A config-verified comparison of how 10 open MoE models route tokens: total and active parameters, expert counts, top-k, shared experts, gating functions and load-balancing methods. 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:MoE Routing Strategies Explained: Top-K, Expert-Choice, Sigmoid Gating and Dynamic Routing (…

原文摘要:MoE routing strategies decide which experts process each token. This guide explains 7 methods, from Mixtral's top-2 softmax to DeepSeek-V4's hash routing, with the models that use 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Nous Research Releases Arete-27B that Cuts Qwen3.8-27B Propaganda From 22% to 0.9%

原文摘要:Nous Research released Arete-27B, an Apache 2.0 fine-tune of Qwen3.8-27B. On Nous’s 803-prompt eval, propaganda answers fell from 22% to 0.9% while Hermes Bench stayed flat at 65.6 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 07 日 2026-11-07 快讯
MarkTechPost 官方资讯

MarkTechPost:Buoyant Releases Kinference: An 开源 Inference Control Plane for Kubernetes and vLLM

原文摘要:Buoyant has released Kinference, an Apache 2.0 inference control plane for vLLM on Kubernetes. It adds declarative routing policies, KV-cache-aware load balancing and per-tenant me 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 06 日 2026-11-06 快讯
11 月 05 日 2026-11-05 快讯
MarkTechPost 官方资讯

MarkTechPost:Baidu Releases ERNIE 5.5 Flash Base: A 280B MoE Base Model With 14B Active Parameters

原文摘要:Baidu has released ERNIE 5.5 Flash Base, a 280B sparse MoE base model with 14B active parameters, trained on 33.5T tokens and published before any mid- or post-training. The post B 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:OpenAI Launches GPT-Image-2.6 Flare and Sunburst: Adds GIFs, SVG Export and 6-Second API Gen…

原文摘要:OpenAI released Images 2.6 and the GPT-Image-2.6 Flare and Sunburst API models. Flare generates images in as little as 6 seconds, and both add GIF output, SVG export, markup edits 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Together AI Releases Together Compiler: Agents Make Inference Up to 2.6x Faster

原文摘要:Together AI released Together Compiler, an agent-driven loop that keeps re-optimizing the inference forward pass as traffic shifts. It beat the best InferenceX configurations by up 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Google AI Research Releases Agile Deliberation Framework for Gray-Area Image Classifiers

原文摘要:Google Research released Agile Deliberation, a human-in-the-loop framework that helps users define subjective visual concepts through borderline images. In 18 live sessions, it lif 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 04 日 2026-11-04 快讯
MarkTechPost 官方资讯

MarkTechPost:Ricursive AI Releases Ripple-65B Preview: A 65B Dense Open-Weight Model Scoring 80.5 on Term…

原文摘要:Ricursive Intelligence released Ripple-65B Preview, a 65B dense open-weight LLM under Apache 2.0. It scores 80.5 on Terminal-Bench 2.1, ahead of US open models up to 15x its size. 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 03 日 2026-11-03 快讯
MarkTechPost 官方资讯

MarkTechPost:Anthropic Releases FlashProve: Cryptographic Proofs That a Reply Came From Committed Weights

原文摘要:Anthropic released FlashProve, an Apache 2.0 zkSNARK prototype that proves a reply came from a committed model without revealing weights. It proves Llama 4 Maverick at 110 tokens/s 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:NVIDIA Shows 3 Ways to Speed Up MoE Inference on DGX Spark, Lifting Ling-3.0-flash Throughpu…

原文摘要:NVIDIA shows how 3 stacked optimizations speed up MoE inference on a single DGX Spark. On Ant Group inclusionAI's 124B Ling-3.0-flash, GSM8K decode throughput rose 42.5% with no de 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Zyphra AI Introduces Zyphra Hierarchical Attention (ZHA): A Path to True Subquadratic Attent…

原文摘要:Zyphra Hierarchical Attention (ZHA) is a tree-indexed sparse attention method that retrofits frozen LLMs with 4B tokens of distillation, matching dense attention on reasoning and 2 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:The Letter Company Releases Letterpress-1: A Coding Agent Built for Growth Engineering

原文摘要:Letter has released the beta of Letterpress-1, a coding agent, harness and IDE for growth engineering. It reads a git-based site and its market, then opens pull requests for SEO, G 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 02 日 2026-11-02 快讯
MarkTechPost 官方资讯

MarkTechPost:NVIDIA Releases Auto Ontology: An Open Semantic Layer That Grounds AI Agent SQL

原文摘要:NVIDIA released Auto Ontology, an Apache 2.0 library that generates a semantic layer from enterprise metadata. It grounds agent SQL in certified definitions and serves context thro 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

11 月 01 日 2026-11-01 快讯
MarkTechPost 官方资讯

MarkTechPost:Nace AI Open-Sources Drex Max v1: A 27B Decision Model That Runs Locally on Apple Silicon

原文摘要:Nace AI open-sources Drex Max v1, a 27B multimodal decision model scoring 67.52 that runs on Apple Silicon. The post Nace AI Open-Sources Drex Max v1: A 27B Decision Model That Run 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 31 日 2026-10-31 快讯
MarkTechPost 官方资讯

MarkTechPost:Meet Factory Droid 1.0: A Model-Independent Coding Agent Harness Built on Script Mode

原文摘要:Factory released Droid 1.0, a model-independent coding agent harness that makes programmatic tool calling the default. On Factory’s own Terminal-Bench 4 runs, Droid scores 65.2% wi 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 30 日 2026-10-30 快讯
MarkTechPost 官方资讯

MarkTechPost:Meet TokenEconomizer: Microsoft’s Plugin That Cuts Coding Agent Costs by Up to 30.9%

原文摘要:Microsoft Research and UC Berkeley released TokenEconomizer, an open-source plugin that learns from past sessions to cut coding agent token costs by 18.4% to 30.9% on SWE-Bench, wi 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Samaya AI Releases Sterling: A 3B-Active Finance Research Agent Model that Beats Claude Fabl…

原文摘要:Samaya AI’s Sterling is a post-trained Nemotron 3.5 Lightning model for financial research. It scores 58.9% on FrontierFinance at $0.61 per question, beating Claude Fable 5.1’s 55. 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 29 日 2026-10-29 快讯
MarkTechPost 官方资讯

MarkTechPost:Prime Intellect Releases Prime Mega MoE: Trainable Fused MoE Kernel, 3.11x Faster Forward Pa…

原文摘要:Prime Intellect released Prime Mega MoE, a backward pass for DeepSeek’s fused Mega MoE kernel. It makes the kernel trainable and speeds Qwen3 MoE training steps by up to 34% in pri 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:AssemblyAI Ships AssemblyAI-TTS-1: A $15 per 1M Character Text-to-Speech API for Voice Agent…

原文摘要:AssemblyAI has released AssemblyAI-TTS-1, its first text-to-speech model for voice agents. It costs $15 per 1M characters, offers unlimited concurrency without a contract, and supp 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Infini-AI-Lab Releases Hermit, an Open-Source Harness Adapter Framework for AI Agents

原文摘要:Infini-AI-Lab released Hermit, an Apache-2.0 harness adapter framework that trains an external layer of instructions, skills, memory and tools from task feedback, leaving the agent 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:NVIDIA Releases Nemotron 3.5 Transcribe: An Open 30B MoE Model for 40-Language Transcription…

原文摘要:NVIDIA released Nemotron 3.5 Transcribe, an open-weights speech LLM with 4.3B active parameters. It transcribes 40 languages, adds speaker labels and word timestamps, and posts a 4 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 28 日 2026-10-28 快讯
MarkTechPost 官方资讯

MarkTechPost:Google DeepMind Introduces Gemini 4 Carbon, Its Top Model for Cyber Defense and Coding

原文摘要:Google DeepMind introduced Gemini 4 Carbon, a frontier model that scores 59 on the Artificial Analysis Intelligence Index and 88.2% on SEC-bench Pro. Access starts with vetted cybe 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Polymathic AI and Princeton Researchers Release FlashMap, a Flow Map Drafter for 6.14x Faste…

原文摘要:FlashMap is a speculative decoding drafter from Polymathic AI, Princeton and USC ISI researchers. It drafts 16 tokens jointly in one pass and reaches up to 6.14x lossless speedup o 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 27 日 2026-10-27 快讯
MarkTechPost 官方资讯

MarkTechPost:OpenAI Launches GPT-6.1 Terra, Halving GPT-6.1 Sol’s API Price for Agent Work

原文摘要:OpenAI released GPT-6.1 Terra, a fast computer-use and vision model priced at $1/$5 per million tokens. It matches GPT-6 Astra’s medium-effort Agents’ Last Exam score at $0.29 per 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 26 日 2026-10-26 快讯
MarkTechPost 官方资讯

MarkTechPost:Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search C…

原文摘要:Nace AI has released Drex for Search, bringing its Drex decision models to document search for AI agents. Nace reports 92.05% on BrowseComp-Plus with 12.6 search calls, and fewer a 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 25 日 2026-10-25 快讯
MarkTechPost 官方资讯

MarkTechPost:Hugging Face Releases Carbon-X, a 14B Open DNA Language Model that Brings DNA and Natural La…

原文摘要:Hugging Face released Carbon-X, an open 14.8B DNA language model built on Qwen3-14B. It reads raw DNA in chat, answers in English, and averages 66.7% across 5 DNA 评测 under 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Researchers from Stanford and UC Berkeley Introduce TokenGate: Token-Gated Attention Makes L…

原文摘要:Stanford and UC Berkeley researchers released TokenGate, a drop-in learned token pruning module for LLMs. It cuts long-context prefill time-to-first-token by 1.4x to 11.3x and peak 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Cognition Open-Sources an Agent Plugin that Helps Secure Systems against Recent Cybersecurit…

原文摘要:Cognition released the Devin Vulnerability Fixes plugin, 3 MIT-licensed agent skills that inventory, patch and test PHP, Node.js and self-managed GitLab deployments hit by this wee 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

10 月 24 日 2026-10-24 快讯
MarkTechPost 官方资讯

MarkTechPost:Meet Anthropic’s sandbox-runtime: OS-Level Sandboxing for AI Agents, No Containers

原文摘要:Anthropic sandbox-runtime (srt) is the open-source, OS-level sandbox behind Claude Code. It restricts file and network access for agents, MCP servers and CLI tools with Seatbelt on 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:Stateful Agentic Engineering Copilots with Letta: Core Memory, Shared State, Semantic Retrie…

原文摘要:Discover how to extend agentic memory beyond basic prompts by building a stateful engineering copilot with Letta. This comprehensive guide covers inspectable core memory, archival 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。

MarkTechPost 官方资讯

MarkTechPost:NVIDIA Releases Nemotron 3.5 Bolt E4B 4GB: An Open Small Language Model for Local Voice Agen…

原文摘要:NVIDIA released Nemotron 3.5 Bolt E4B 4GB, an open small language model post-trained from Gemma 4 E4B for local voice assistants, game characters and robots, targeting about 4GB of 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。