Research

The AllSpark team has released Iris-mini and Iris-pro, two open-source search agents built on Qwen models that lead benchmarks among open-weight models in their size classes.

In a new robotics benchmark, GPT-6 Astra shows major gains in spatial understanding. On StationeryBench, the model completed 7 out of 100 tasks with dual-arm robots, while competitor MolmoAct2 couldn't finish a single one.

Anthropic launches Claude Fable 5.1 and Mythos 5.1, its most capable AI models yet. Fable 5.1 doubles its predecessor's score on Terminal-Bench-Science and improves agentic coding by over 30 percent.

Google Deepmind is testing a double-blind evaluation of a frontier AI model for the first time. Cryptographic protection through Confidential Space is meant to keep Google from seeing the test questions and keep evaluators from seeing the model weights.

Rob Strechay, until recently managing director and principal analyst at theCUBE Research, has joined VentureBeat as our first Lead Analyst and a founding analyst of VentureBeat Research.

Artificial Analysis has released the "Search Index," a benchmark that rates search API providers for AI agents on quality, cost, and speed. Of seven providers tested with GPT-5.6 Luna, Parallel, Exa, and Firecrawl scored highest.

Google shipped Gemini 3.7 Flash just three weeks after 3.6 Flash. The new model is supposed to be Google's most capable workhorse yet for coding and AI agents, and according to the company's own benchmarks, it beats Claude Sonnet 5 and GPT-5.

Tool-using language-model agents are governed not only by task prompts but also by persistent system-side instructions that specify tools, arguments, policies, execution protocols, and recovery.

As large language models move from isolated task solving toward long-term service in human environments, they require social intelligence: the ability to infer mental states, track social relations, reason over norms, and adapt behavior under context.

Anthropic's new flagship model Claude Opus 5 posts top scores in coding and knowledge work at half of Fable 5's token rates. On ARC-AGI-3, a benchmark for novel problem-solving, Opus 5 hits 30.2 percent, nearly four times higher than GPT-5.6 Sol.

Users frequently express their beliefs to large language models (LLMs). In some situations, the LLM should accept these contextual beliefs as true. In others, they should stick to their prior knowledge.

Kimi is launching K3, a multimodal open-weight model with 2.8 trillion parameters and one million tokens of context. In the company's own benchmarks, it comes close to Claude Fable 5 and GPT 5.6 Sol while beating Opus 4.8 and GLM 5.

Across 107 enterprises, AI infrastructure spending is accelerating well ahead of the ability to see or steer its economics. Most organizations run their AI on a familiar base of hyperscalers and model-provider APIs, yet the next dollar is aimed at specialized...

xAI releases Grok 4.5, trained on tens of thousands of Nvidia GB300 GPUs. In coding benchmarks, the model trails Fable 5 and GPT-5.5 but needs 4.2 times fewer tokens than Opus 4.8. At $2 per million input tokens, it costs a fraction of the competition.

As AI coding agents become more autonomous, they increasingly ship code iteratively, with the codebase persisting across sessions. This persistence creates a new attack surface: a misaligned or prompt-injected agent can distribute attacks across pull requests...

Anthropic released Claude Sonnet 5, which beats its predecessor Sonnet 4.6 across all benchmarks and even edges past the larger Opus 4.8 on the GDPval-AA v2 knowledge work test with a score of 1,618.

Anthropic released Claude Science, an AI workbench for researchers. More than 60 preconfigured skills cover fields like genomics and computational chemistry, and a verification agent automatically checks citations and calculations.

Agentic language models dramatically expand the applications of AI yet little is publicly known about how to curate training data for broadly capable agents.

OpenAI is expanding its Daybreak cybersecurity initiative with an updated Codex Security plugin, the full GPT-5.5-Cyber model, and a partner network with more than 25 security firms and several governments.

For a quarter century, the Google search box has been one of the most recognizable interfaces in computing: a thin white rectangle, a blinking cursor, a few typed words, and a list of blue links. On Tuesday, Google will formally retire that paradigm.

Railway , a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence app...

The artificial intelligence coding revolution comes with a catch: it's expensive. Claude Code , Anthropic's terminal-based AI agent that can write, debug, and deploy code autonomously, has captured the imagination of software developers worldwide.

Alfred Wahlforss was running out of options. His startup, Listen Labs , needed to hire over 100 engineers, but competing against Mark Zuckerberg's $100 million offers seemed impossible.

Salesforce on Tuesday launched an entirely rebuilt version of Slackbot , the company's workplace assistant, transforming it from a simple notification tool into what executives describe as a fully powered AI agent capable of searching enterprise data, drafting...

Anthropic released Cowork on Monday, a new AI agent capability that extends the power of its wildly successful Claude Code tool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, la...

Nous Research , the open-source artificial intelligence startup backed by crypto venture firm Paradigm , released a new competitive programming model on Monday that it says matches or exceeds several larger proprietary systems — trained in just four days using...

Tech Report GitHub Hugging Face ModelScope DISCORD Introduction We are excited to introduce Qwen3Guard, the first safety guardrail model in the Qwen family.

Three security researchers used Anthropic's Claude models to break into OpenAI's internal systems through its community forum in less than 72 hours. According to the team, Opus 5 succeeded where its predecessor couldn't bypass a common security measure.

César de la Fuente’s lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections.

Introducing ChatGPT for Financial Services, combining built-in financial data and GPT-6 Astra for research, modeling, and client-ready materials.

Anthropic is launching an API that lets regulators, media outlets, and researchers check whether text carries Claude's digital watermark. The EU AI Act now requires invisible watermarks in AI-generated text.

The Pew Research Center analyzed nearly half a million English-language web pages for AI-generated content. More than a third of pages published since ChatGPT's launch show signs of machine-written text, and commercial .

AI companies like Anthropic and OpenAI regularly publish reports on how people are using products like Claude and ChatGPT, but they only release the data they want us to see, AI researchers say.

Explore new education plugins for ChatGPT Work and Codex that help K–12 teachers, college educators, and students learn, teach, research, and build.

New OpenAI research shows how AI is expanding what workers do, with ChatGPT users taking on tasks across roles and reshaping job boundaries.

Limited memory language models (LMLMs) externalize factual knowledge during pretraining to a knowledge base (KB), rather than memorizing it in their weights. During generation, the model then fetches knowledge from the KB as needed.

At an event for pharmaceutical executives, biotech founders, and researchers on Tuesday, Anthropic announced Claude Science, a major new product intended to support scientific research in the same way that Claude Code supports software engineering.

Asynchronous monitoring, incident investigations, and compliance audits primarily rely on agent traces to reconstruct what happened. These analyses assume that LLM agents cannot tamper with their own execution traces.

MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses across realistic mental health conversations.

Scene safety understanding plays a life-or-death role in situational awareness in various critical domains. Traditional methods that rely on learning direct mappings between scenes and safety levels often lack interpretability, limiting their reliability in cr...

Bayesian filtering provides a principled framework for online state estimation under uncertainty, yet its application to systems with high-dimensional states and complicated posterior distributions remains challenging.

Qwen3.8-Omni-Flash is Qwen's first multimodal model designed for AI agents. It processes audio and video together and independently uses tools to edit vlogs, translate clips, or summarize movies. On audio-video benchmarks, it nearly matches Gemini 3.

Fallacy-detection benchmarks pair fallacy classes with a single "valid" or "none" class that takes everything data collection did not label as a fallacy.

Thermal optimization of 2D CFET inverters requires testing structural proposals against their electrical costs. We examine these research tasks using an AI agent workflow within a supplied electrothermal model.

Google Deepmind has expanded Co-Scientist from a hypothesis generator into a research system that's integrated into the lab. Across three disciplines, from materials synthesis to the autonomous development of a medical AI architecture, the Gemini-based multi-a...

Factual question answering (QA) typically assumes a single canonical answer, obscuring whether large language models (LLMs) retain divergent accounts of long-tail facts.

Long-horizon agentic tasks require large language models (LLMs) to iteratively retrieve, integrate, and maintain dispersed information across multi-turn interactions, but preserving all interaction histories leads to a continuously growing working context.

Agent skills package specialized knowledge and workflows into reusable resources that extend AI agent capabilities. Recent work automatically discovers such skills from agent experience, which enables agents to progressively adapt through interaction.

OpenAI research reveals how enterprises are adopting agentic AI, using ChatGPT and Codex, and how frontier firms are pulling ahead in AI adoption.

Thinking Machines, the AI lab from former OpenAI CTO Mira Murati, has released Inkling Small. The open-weights reasoning model is less than a third the size of Inkling but beats it on several coding and reasoning benchmarks.

Long visual context poses a challenge for vision-language models: performance degrades as the number of distractors grows, and processing all tokens at once is computationally infeasible under GPU memory constraints.

On-policy distillation provides dense supervision for multimodal reasoners, but its trajectory-level reward cannot determine whether a failed answer arose from perception or subsequent reasoning.

How two API settings improved GPT-5.6 performance on ARC-AGI-3, boosting scores and efficiency by retaining reasoning and enabling compaction.

We introduce Shieldstral, a 3B-parameter policy-adaptive multimodal safety classifier that matches or outperforms models nearly 7$\times$ its size on text safety benchmarks and sets a new state of the art on multimodal safety classification.

Theory of Mind (ToM), the ability to infer other's beliefs, intentions, and states of knowledge, is central to social interaction, yet remains challenging for current Multimodal Large Language Models (MLLMs), especially in multi-party meetings where cues are d...

Across 107 enterprises, AI agents are being given real access to systems and data while the controls meant to contain them lag behind. More than half have already had a confirmed agent security incident or a near-miss; only about a third give every agent its o...

Recent robot foundation models operate with single-step or short-history visuomotor context. We introduce Test-Time-Training Robot Policies (RoboTTT), a robot model and training recipe that scale visuomotor context to 8K timesteps, three orders of magnitude be...

Real repository issues routinely include visual evidence such as screenshots, error dialogs, rendered UI states, and logs, yet repository-level issue localization is evaluated mostly as a text-only task.

Long-term memory has become a foundational capability for LLM-based agents that accompany users across extended, multi-session interactions. Existing benchmarks, however, evaluate such memory almost exclusively through downstream question answering, scoring on...

According to OpenAI, GPT-5.6 Sol independently fine-tuned the smaller Luna model, triggered by a single "fairly under-specified prompt." In OpenAI's internal RSI benchmark for recursive self-improvement, Sol scores 16.2 points higher than GPT-5.5.

The rapid development of large language models and multimodal large language models has accelerated the emergence of proactive agents capable of operating everyday tools and assisting users in real-world environments.

Reasoning has become a core capability for large models, especially when reliable decisions require understanding logical consequences. Recent video generation models offer a reasoning path distinct from previous Chain-of-Thought (CoT): reasoning can unfold th...

Structure-property relationships are foundational to biology, chemistry and materials science, where function, reactivity and physical response emerge from spatial, chemical and periodic organization.

A new analysis from OpenAI reveals issues in SWE-Bench Pro, a popular coding benchmark, raising concerns about reliability and accuracy in evaluating AI models.

LLMs memorize sensitive training data, including personally identifiable information (PII), creating a pressing need for reliable post hoc removal methods.

The evaluation of long-term video quality understanding remains an open challenge for large vision-language models (LVLMs). Existing video quality benchmarks predominantly focus on short clips and isolated distortions, overlooking the temporal continuity, cumu...

Certified training aims to produce models whose predictions can be formally verified against adversarial perturbations, typically by optimising upper bounds on the worst-case loss over an allowed perturbation set.

Introducing GeneBench-Pro, a new benchmark testing AI performance in genomics, biology, and scientific research using complex, real-world datasets.

In the big-world regime, agents cannot be universally capable and their ability is inevitably specialized across a world model in pieces. Consequently, standard uniform guarantees fail to distinguish between the understanding of critical bottlenecks and irrele...

Japanese AI startup Sakana AI is launching Fugu, a system that coordinates multiple AI models on the fly to compete with leaders like Anthropic's Fable 5. The approach also aims to cut dependence on any single AI provider.

Data Management

Latent world models are typically trained to predict factual transitions, whereas model predictive control (MPC) must compare alternative actions from the same state.

Effective summarization of interrogative dialogues is a critical task in forensic and investigative settings, requiring high factual accuracy, coherence, and attribute-specific relevance.

Real-world time series evolve continuously, with meaningful changes potentially emerging at any moment. However, existing time-series language models (TSLMs) remain inherently static.

Empirical evaluation is central to recommender-systems (RecSys) research, but turning experimental designs into executable code remains a manual and error-prone task.

We are expanding our AI & Economy team with world-class academic advisors, fellows, and core internal researchers.

Characterizing the physical properties of clay and cementitious materials matters across many fields, from materials science to geological waste disposal.

New OpenAI Economic Research shows how workers use AI beyond traditional roles and which new activities become recurring parts of their work.

Large language models can easily generate explanations for their own outputs, but such self-explanations are not necessarily faithful to the model's behavior.

See how an MIT researcher uses GPT-5.6 Sol with Codex to autonomously run quantum computing experiments, analyze results, and calibrate qubits.

Many recurring text functions are easy to describe but difficult to implement with rules, while calling a large remote model for every input introduces repeated cost, latency, and dependency on a provider.

Text-rich visual inputs require models that can read, retrieve, and compress language directly in pixel space, yet existing pixel-text encoders struggle with fixed resolution pretraining, visual shortcut learning, weak visual grounding, and multilingual visual...

ChatGPT can now connect to trusted healthcare data, helping clinicians securely access patient context, medical research, and more.

Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification.

Google introduces AMIE for real-time clinical video consultations in simulated settings.

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Last week, I headed 30 miles south of San Francisco to a hotel in Mountain View, California, to join some of the most acco...

Model ML uses GPT-5.6 Sol to carry finance work from research and analysis through editable, traceable PowerPoint decks and Excel workbooks.

Meet GPT-5.6-Cyber, OpenAI’s cybersecurity-specific model available through Daybreak Red for authorized vulnerability research, exploit validation, and security testing.

Virgin Atlantic is accelerating research, product planning, and decision-making with ChatGPT Work, helping teams connect signals across the customer journey.

OpenAI explains recent third-party cybersecurity evaluation incidents and outlines new safeguards to strengthen AI model testing and evaluation.

Depth-routing residual architectures allow Transformer layers to retrieve earlier representations instead of inheriting only the immediately preceding state.

OpenAI is giving 100,000 academic researchers free access to ChatGPT's most advanced AI models to accelerate scientific research, collaboration, and discovery.

Amazon is scaling back most of its in-house Nova AI models, including Nova Premier, Omni, Reel, and Canvas. The models stay online for existing customers in "keep the lights on" mode but are no longer actively developed.

Quantum error correction (QEC) is indispensable for scalable fault-tolerant quantum computing. However, discovering QEC codes that remain effective is challenging, as logical performance depends on the interplay between code structure, hardware, syndrome extra...

Controllable video generation remains challenging due to the difficulty of specifying precise multi-object interactions using text prompts or motion-control inputs that primarily constrain pixel movement.

A field experiment with 1,559 Pakistani judges found that the AI assistant JudgeGPT boosted case resolution by 6.3 percent. The catch: only judges who got hands-on training saw gains. Without it, the effect mostly disappeared.

The next time you apply for a job, AI may screen your résumé before any human sees it. But there’s good reason to question whether AI will judge you fairly. Researchers already know that LLMs pick up human biases from their training data.