The frontier model race

Liquid AI released LFM2.5-24B-A2B, a reasoning model with 24 billion total parameters, 2 billion active per token and a 131,072-token context window. The company reports 184 tokens per second on an Apple M5 Max and stronger results than larger rivals on several tool-use tests, while acknowledging weaker coding and knowledge scores. Downloadable weights and support for MLX and llama.cpp make it a practical option for private agents running locally.

Liquid AI ↗ · October 14 briefing · Open models

AI hardware startup Natura unveiled Interface, a $99 smart ring that lets users summon AI agents such as Claude, ChatGPT, Meta's Muse and Grok Bot with a press of a finger. The ring also tracks heart rate, sleep and skin temperature, records meetings and controls smart home devices, with 6 to 12 days of battery life. Preorders open in November and shipping starts in December or January, with a $9 monthly subscription after a free trial.

TechMeme ↗ · October 8 briefing

Anthropic released Claude Haiku 5.5, pricing it at $0.10 per million input tokens and $0.50 per million output tokens for requests under 100,000 tokens, a 90% cut that matches OpenAI's GPT-6 Luna. The company positions Haiku as a supporting worker for Opus and Sonnet, handling summarization, classification and database queries. It scored 1,620 on the GDPval-AA knowledge work benchmark versus 735 for Haiku 4.5, and is available on AWS, Google Cloud and Azure.

VentureBeat ↗ · October 7 briefing

OpenAI is rolling out GPT-6 across ChatGPT alongside Intelligent UI, a feature that lets responses mix text with diagrams, charts, forms, calculators and tappable buttons. Paid Plus, Pro, Business and Enterprise users get GPT-6 Sol starting October 7, while Free and Go users receive GPT-6 Luna from October 8. ChatGPT can still default to plain text when that fits better, and the update reaches more than 1.2 billion weekly users.

The Verge ↗ · October 7 briefing

French lab Mistral released Mistral Large 4, nicknamed Le Chonk, a multimodal mixture of experts model with 1.05 trillion total and 49 billion active parameters. It is in public preview through Mistral's API, with open weights planned by the end of October after safety testing. Mistral says it trained the model on about 3,800 Nvidia Grace Blackwell GPUs in its own European data centers, far less compute than Chinese rivals.

TechCrunch AI ↗ · October 6 briefing · Open models

Google DeepMind launched EmbeddingGemma 2, a 740 million parameter open model that maps text, code, images, video and audio into a single shared embedding space. Released under the Apache 2.0 license, it is built on Gemma 4 and designed to run on consumer hardware. Google calls it its most capable model for on-device multimodal search and retrieval.

Unite. AI ↗ · October 6 briefing · Open models

Reflection AI, founded by two former Google DeepMind researchers, released Beam, a 501 billion parameter open-weight model it pitches as America's answer to Chinese open-source AI. The company says Beam excels at coding and agentic tasks, scores close to leading Chinese open models, and runs on three to four times less compute than comparable Western open models. Valued at $25 billion, Reflection has a $6.3 billion compute deal with SpaceX, supply agreements with the US Department of Energy and Department of War, and is already training a larger successor.

Fortune AI ↗ · October 5 briefing · Open models

Anthropic released Sonnet 5.5, which it says runs about 30 percent faster and consumes significantly fewer tokens than Sonnet 5, released three months earlier. Despite sitting below the flagship Opus line, Sonnet 5.5 outperforms Opus 5.5 on agentic coding tasks because it can run multiple agents without exceeding budget constraints, and it now matches Opus 5's cybersecurity safeguards. Anthropic says it plans to release an updated Haiku model in the coming weeks, continuing a rapid release cadence as labs compete on cost and speed.

TechCrunch AI ↗ · September 28 briefing

SpaceXAI, formed when xAI merged with SpaceX, released Grok 4.7 with unchanged pricing of $2 per million input tokens and $6 per million output tokens. The model improved to 71.0 percent on a leading software engineering benchmark, up from 65.2 percent for Grok 4.6, and scored 46.3 percent on a separate coding test, ahead of GPT-5.6 Sol Max but behind Claude. On the tougher Terminal-Bench 4.0 agentic coding test, Grok 4.7 scored just 26 percent, well behind GPT-6 Astra and Claude.

Next Big Future AI ↗ · September 21 briefing · AI that writes code

Anthropic launched Claude Code Projects in beta, a coordination layer that splits a coding task into parallel cloud sessions that keep running after a developer closes their laptop. The system delegates work across threads, reviews outputs, and assembles a finished result, while keeping a shared memory so it can recall details like a delayed release date without re explaining context. The feature is available to select Claude Pro and Max subscribers in Claude Code.

Anthropic ↗ · September 17 briefing · AI that writes code

Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its most advanced live dialogue models yet, aimed at powering production-grade voice agents. The models execute tools and API calls in the background without interrupting conversation, process live visual input, and switch between 97 languages mid conversation. Extended Thinking ranks first on Artificial Analysis' Speech to Speech Quality Index, and both are now available via the Gemini API.

AI Research News ↗ · September 15 briefing

Cognition, the company behind the Devin coding agent, released SWE-2, its most capable coding model to date, post-trained with reinforcement learning on top of Moonshot AI's 2.8 trillion parameter open model, Kimi K3. On a leading coding benchmark, SWE-2 scored within one point of Fable 5.1 while costing 64 percent less to run. The release signals a broader trend of startups fine-tuning open weight Chinese models to compete with closed frontier labs on price.

AI Research News ↗ · September 13 briefing · Open models · AI that writes code

OpenAI officially released GPT-6 Astra, a frontier model the company claims marks the onset of artificial general intelligence (AGI) with capabilities that outperform humans on most economically valuable work. The model can navigate software autonomously without user input, working across browsers, spreadsheets, applications, and producing finished documents and workflows rather than just providing instructions. OpenAI co-founder Greg Brockman announced: "Welcome to the AGI era," signaling what the company views as a fundamental inflection point in AI development.

VentureBeat ↗ · September 3 briefing

OpenAI has opened a preview of Ultrafast, a new API service tier that runs its flagship GPT-5.6 Sol model at up to 14 times normal speed, delivering as many as 750 output tokens per second. The tier is powered by Cerebras wafer scale hardware rather than conventional GPU clusters, a speed no GPU cloud has publicly matched. OpenAI is aiming the mode at enterprise workloads where latency matters most, naming incident response, customer support, financial analysis and e commerce as target use cases. Access is limited for now, and the company says it will widen availability as capacity grows.

TechCrunch AI ↗ · August 13 briefing

SpaceXAI, the company formerly known as xAI, released Grok 4.6 on August 12, a new flagship model tuned for long-running agents and more ambitious interactive and visual work. The company says it scores 61 on the Artificial Analysis Intelligence Index, matching GPT-5.6 Sol and trailing only Opus 5 and Fable 5, with standout results on the GDPVal and coding benchmarks. Pricing starts at $2 per million input tokens and $6 per million output tokens, which xAI pitches as about half the price of other frontier models. It shipped the same day in the Cursor code editor, the company's Grok Build tool, and its API.

Unite. AI ↗ · August 12 briefing

DeepSeek released the production build of its flagship model, DeepSeek V4 Pro, ending a preview that ran nearly four months. The version, designated V4 Pro 0813, appeared on August 12 as the general availability release on OpenRouter's model page, and DeepSeek's own API documentation now lists it behind the deepseek-v4-pro endpoint. Pricing carries over unchanged from the preview at $0.435 per million input tokens on a cache miss, keeping it among the cheapest frontier class models available.

Unite. AI ↗ · August 12 briefing

Nvidia released Nemotron 3.5 Lightning, a 30 billion parameter mixture of experts model that companies can download, modify and deploy for free without asking permission. Nvidia claims it delivers up to four times faster output and 30 percent faster task completion than comparable models in its class, and it runs on a single PC GPU including RTX PCs and DGX Spark systems as well as across data centers. Alongside it, Nvidia open sourced Switchyard, a model routing library that automatically directs each step of an agent workflow to the cheapest suitable model. CrowdStrike, CodeRabbit and Harvey were among the companies that tested and customized the model ahead of launch.

CNBC Technology ↗ · August 11 briefing · Open models

xAI released Grok Imagine Image 2.0, making it the Quality Mode on the company's Imagine site and in its iOS and Android apps. The model is built around editing, adding a magic wand for regional changes, segmentation for precise selection, background removal with transparent export, and smart resize across nine aspect ratios, and it accepts up to five reference images in a single generation. xAI says it trained the model for fidelity across photography, design and illustration, with extra attention to typography and layout rendering. By the company's own figures it ranks second globally in both text to image generation and image editing, behind OpenAI's gpt-image-2, with API access still to come.

Unite. AI ↗ · August 9 briefing

Microsoft unveiled MAI-Cyber-1-Flash, its first AI model built specifically for cybersecurity, embedded inside MDASH, the company's multi-agent system for finding and fixing software vulnerabilities. Microsoft says the configuration scores 96% on a leading industry benchmark, about 12 points above a rival frontier model, while cutting costs by nearly half, with the compact model handling up to 90% of queries and handing the hardest 10% to the larger GPT-5.4. The company also introduced Project Perception, an agentic security system entering public preview on August 3, advancing its argument that the future favors the cheapest capable model routed intelligently rather than the biggest one.

VentureBeat ↗ · July 27 briefing

Moonshot AI, the Beijing-based startup backed by Alibaba, released Kimi K3, a 2.8-trillion-parameter open-source model that the company says is the largest ever released, with benchmarks showing it rivals top proprietary systems from Anthropic and OpenAI. The model features a 1-million-token context window, native visual understanding, and an always-on reasoning mode, with full weights scheduled for release on July 27. On benchmarks such as GDPval-AA v2 and AA-Briefcase, K3 placed third and second respectively, while topping the BrowseComp long-horizon information-seeking benchmark and ranking first on Arena. AI's Frontend Code Arena.

VentureBeat ↗ · July 16 briefing · Open models

SpaceXAI has unveiled Grok 4.5, its first joint AI model built with newly acquired coding startup Cursor, in a bid by Elon Musk's firm to gain ground on rivals like Anthropic and OpenAI. Unlike earlier Grok versions built for general chatbot use, Grok 4.5 is designed mainly for legal work such as reviewing contracts and case files, financial tasks like statement analysis and due diligence, and software engineering across large codebases. SpaceXAI is pricing the model at 2 dollars per million input tokens and 6 dollars per million output tokens, and is rolling it out across Cursor's desktop, web, and command line platforms, though it is not yet available in the EU.

Bloomberg Technology ↗ · July 8 briefing · AI that writes code

Mistral has launched Robostral Navigate, a new hardware agnostic robotics navigation model, as the French startup expands into the emerging field of physical artificial intelligence. The model is trained entirely via simulation and needs only a single camera along with basic language prompts to guide a robot's movement, rather than relying on expensive specialized sensor rigs. The release positions Mistral to compete in the growing race among AI labs to bring foundation models into real world robotics applications.

TechMeme ↗ · July 8 briefing · Robots and physical AI

SpaceXAI and Cursor plan to launch their first jointly developed AI model as soon as Wednesday, according to an internal memo sent to staff and reported by The Information. The companies pushed back the original launch timeline this week to improve the model's efficiency before release. The new model follows SpaceX's all-stock acquisition of Cursor, part of a push by Elon Musk's company to build what it calls the world's most useful AI models. It is expected to roll out inside Cursor's coding tools and xAI's Grok Build platform.

TechMeme ↗ · July 7 briefing · AI that writes code

Z.ai, the Beijing-based artificial intelligence lab formerly known as Zhipu AI, on Wednesday officially launched ZCode, a free desktop application it describes as an "Agentic Development Environment" purpose-built for its flagship GLM-5.2 large language model. The move marks the company's most aggressive push yet into the fast-growing AI-powered coding tool market, where it now competes directly with Cursor, Claude Code, GitHub Copilot, and Google's Antigravity."Introducing ZCode, the official development environment for GLM-5.2," the company wrote on X, noting the tool is available on macOS, Windows, and Linux, supports bring-your-own-key (BYOK) configurations for third-party models, and offers a 1.5x usage-quota bonus for subscribers to its GLM Coding Plan. Read one way, ZCode is simply another entrant in a crowded market. Read another, it is a single product that crystallizes three of the most consequential trends in enterprise software today: the race-to-the-bottom pricing of frontier AI models, the geopolitical balkanization of the AI stack, and the rapid maturation of agentic coding agents into what Gartner now estimates is a roughly $10 billion market. An AI coding tool designed to think in projects, not promptsUnlike traditional IDEs that bolt on AI through a chat sidebar or autocomplete extension, ZCode is best understood as an agent-first development environment. Its core design is built around long-horizon tasks: the user describes an ou...

VentureBeat ↗ · July 2 briefing · AI that writes code

Rebecca Bellan / TechCrunch: Anthropic launches Claude Science, an AI workbench that uses existing Claude models like Opus 4.8 to integrate 60+ scientific databases and specialized toolkits  —  Anthropic introduced Claude Science on Tuesday, an AI workbench that gives scientists one environment to do computational research …

TechMeme ↗ · June 30 briefing

Anthropic’s Claude Sonnet 5 brings stronger agentic capabilities, lower pricing, and improved safety, positioning the model as a cheaper alternative to Opus, GPT-5.5, and Gemini Pro.

TechCrunch AI ↗ · June 30 briefing

New models are launching in Asia that promise Mythos-like capabilities without fear of an export ban. U. S. AI labs may never recover this enormous market.

TechCrunch AI ↗ · June 28 briefing · Rules and the courts

OpenAI is announcing a limited preview of its newest frontier AI model GPT-5.6 family, which comes in three variants: Sol, Terra, and Luna. Sol is for the hardest problems, such as complex coding and security research; Terra is for high-volume business tasks like customer support, internal tools and document analysis; and Luna is for faster, lower-cost everyday work like summarization, drafting and routine automation. Sol and Terra set new high benchmark scores, while Luna performs near GPT-5.5 levels on several tests despite being positioned as the fastest and lowest-cost model in the GPT-5.6 family. However, the models are being made available initially to a narrow set of approximately 20 total organizations, after OpenAI shared the models and release plans with the U. S. government. A general release is planned for "the coming weeks."The staggered release follows an executive order issued by President Donald J. Trump earlier this month on June 2, 2026, which calls upon various federal agencies to collaborate on a process for benchmarking and assessing capabilities of new AI models to ensure they are safe and appropriate for wide release. While this process remains underway (it was said in the order to take 30 days, so July 2), OpenAI says in its release blog post that it "previewed our plans and the models’ capabilities ahead of today’s launch. At [the U. S. government's] request, we are starting with a limited preview for a small group of trusted partners...

VentureBeat ↗ · June 28 briefing

TechCrunch AI / VentureBeat / Engadget / TechMeme ↗ · June 23 briefing

Grok Build's in-terminal marketplace bundles skills, agents, hooks, and MCP servers, with commit-SHA verification on every remote plugin. The post xAI Ships Grok Build Plugin Marketplace With MongoDB, Vercel, Sentry, Chrome DevTools, Cloudflare, and Superpowers Plugins at Launch appeared first on MarkTechPost.

AI Research News ↗ · June 11 briefing

Google DeepMind open-sourced DiffusionGemma, a 26B MoE text-diffusion model that generates text up to 4x faster than comparable autoregressive models for on-device AI workloads.

Ars Technica ↗ · June 10 briefing

JetBrains releases Mellum2 under Apache 2.0 — a 12B MoE model trained on 10.6 trillion tokens for AI workflows. The post JetBrains Releases Mellum2: A 12B MoE Model for Fast, Specialized Tasks in Multi-Model AI Pipelines appeared first on MarkTechPost .

Planet AI (Aggregator) ↗ · June 2 briefing · Open models

Big news in enterprise AI broke over the weekend as Chinese AI startup MiniMax released its highly anticipated M3 large language model on Sunday evening Eastern time, pairing frontier-tier coding and agentic performance with a 1-million-token context window and native multimodality for a fraction of the cost of leading proprietary models, with pricing starting at just $20 per month under its new subscription token plans. The company's leadership also announced plans to deliver the model und...

VentureBeat ↗ · June 1 briefing

Microsoft Research introduces Webwright, a terminal-native browser agent framework that replaces click-trace web automation with reusable Playwright scripts. Using a single agent loop across three modules and roughly 1,000 lines of code, Webwright powered by GPT-5.4 reaches 60.1% on the long-horizon Odysseys benchmark and 86.7% on Online-Mind2Web — the highest AutoEval score among open-sourced harness recipes. The post Microsoft Research Releases Webwright: A Terminal-Native Web Agent Framework ...

AI Research News ↗ · May 24 briefing

Russell Brandom /TechCrunch:Adaption, co-founded by ex-Cohere VP of AI research Sara Hooker, unveils AutoScientist, which can automate the research loop behind model training and alignment— For years, AI researchers have anticipated the moment when AI systems will be able to improve themselves better than humans could.

TechMeme ↗ · May 13 briefing

OpenAI is releasing an update to its artificial intelligence image-generating software that it says will let users create accurate, complex charts and scientific diagrams, part of a bid by the company to make its technology more appealing to professionals.

Bloomberg Technology ↗ · April 21 briefing

Moonshot open-sources Kimi K2.6 with claims of stronger long-horizon coding under a modified MIT-style license.

TechMeme ↗ · April 20 briefing · Open models · AI that writes code

Claude Mythos Preview is Anthropic's most powerful AI model that excels at identifying weaknesses and security flaws within software.

CNBC Technology ↗ · April 16 briefing

Launch of GPT-5.4-Cyber follows concern about ability of Anthropic’s Mythos to find software bugs

Financial Times Tech ↗ · April 15 briefing

Google is launching a new Gemini app on Mac that allows you to interact with the AI assistant without switching windows on your desktop. With the app, you can use the Option + Space shortcut to pull up a floating chat bubble, where you can ask Gemini questions and share your window. Before sharing your […]

The Verge ↗ · April 15 briefing

OpenAI is letting a select group of users access a new artificial intelligence model that's meant to be more adept at spotting software security vulnerabilities, one week after rival Anthropic PBC announced a limited release of an AI tool called Mythos.

Bloomberg Technology ↗ · April 14 briefing

Meta Platforms Inc. debuted its latest artificial intelligence model Wednesday — its first since Chief Executive Officer Mark Zuckerberg embarked on a multibillion-dollar overhaul of the company’s AI organization to keep pace with rivals.

Bloomberg Technology ↗ · April 8 briefing

The new offering is designed for advanced reasoning and multimodal capabilities.

AI Business ↗ · April 5 briefing · Open models

Google has released Gemini 3.1 Flash Live in preview for developers through the Gemini Live API in Google AI Studio. The model targets low-latency, natural real-time voice and multimodal streams and is positioned as Google's highest-quality audio and speech model to date, with tool use for agents.

AI Research News ↗ · March 29 briefing

Cohere launched Transcribe, a 2B-parameter open-source automatic speech recognition model for notetaking, speech analysis, and voice workloads.

TechMeme ↗ · March 26 briefing

Harry McCracken / Fast Company: Sundar Pichai and Google executives on being blindsided by ChatGPT, Gemini, and a vision of useful AI everywhere.

TechMeme ↗ · March 23 briefing

Anthropic launched Code Review in Claude Code, a multi-agent system that automatically analyzes AI-generated code, flags logic errors, and helps enterprise developers manage the growing volume of code produced with AI.

TechCrunch ↗ · March 9 briefing · AI that writes code

Microsoft has released a 15 billion parameter open-weight multimodal model designed for image and text tasks requiring perception and selective reasoning. The model excels at scientific and mathematical reasoning and GUI understanding while balancing compute efficiency.

AI Research News ↗ · March 8 briefing · Open models

OpenAI has officially released GPT-5.4, its most capable model for professional work. The update features state-of-the-art coding, native computer use, and a massive 1M-token context window. It includes 'Pro' and 'Thinking' versions designed for complex reasoning and enterprise-grade tool integration.

OpenAI Blog ↗ · March 5 briefing

The AI model wars have reached a new peak as OpenAI quietly launched GPT-5.3 Instant. The new model is designed to be 'less cringe' and more accurate, competing directly with Google's newly released Gemini 3.1 Flash-Lite.

Forward Future / TechCrunch ↗ · March 4 briefing

Google Home has introduced 'Live Search,' a Gemini-powered feature that allows users to search through their smart camera's video history using natural language queries like 'when did the cat jump on the counter?'

Engadget ↗ · March 3 briefing

Google has officially launched agentic task automation for Gemini on the upcoming Pixel 10 and Samsung Galaxy S26. The update allows the AI to autonomously navigate third-party apps like Uber and DoorDash to complete multi-step requests, marking a shift from chatbots to true digital assistants.

TechMeme ↗ · February 26 briefing

Anthropic unveiled Claude Cowork, an agentic platform designed to automate complex office tasks. The system features deep integrations with Google Workspace, Salesforce, and WordPress, allowing AI agents to handle HR, design, and engineering workflows autonomously.

The Verge ↗ · February 25 briefing

Google has officially released Gemini 3.1 Pro, claiming significant improvements in complex problem-solving, long-context reasoning, and agentic task execution.

Ars Technica ↗ · February 20 briefing

Zyphra has released ZUNA, a 380M-parameter foundation model for EEG data, marking a major breakthrough in noninvasive brain-computer interface (BCI) technology.

MarkTechPost ↗ · February 20 briefing

Elon Musk’s xAI has released Grok 4.20 in Beta, featuring a 'multi-agent council' where 4 AI agents collaborate in real-time. The model has a provisional Elo score of ~1535 in early testing.

Next Big Future AI ↗ · February 18 briefing

Anthropic has released Claude Sonnet 4.6, making it the new default model for both free and paid users. The update brings frontier-level coding and reasoning capabilities to the faster Sonnet tier.

CNBC Technology ↗ · February 18 briefing

Alibaba's Qwen team has released Qwen3.5-397B, a Mixture-of-Experts (MoE) model with 17B active parameters and a massive 1M token context window, specifically optimized for long-horizon reasoning in AI agents.

MarkTechPost ↗ · February 17 briefing

OpenAI is releasing its first artificial intelligence model that runs on chips from semiconductor startup Cerebras Systems Inc., part of a push by the ChatGPT maker to broaden the pool of chipmakers it works with beyond Nvidia Corp.

Bloomberg Technology ↗ · February 13 briefing · The compute buildout

OpenAI is releasing its first artificial intelligence model that runs on chips from semiconductor startup Cerebras Systems Inc., part of a push by the ChatGPT maker to broaden the pool of chipmakers it works with beyond Nvidia Corp.

Bloomberg Technology ↗ · February 12 briefing · The compute buildout

Z.ai released GLM-5, claiming best-in-class performance for reasoning, coding, and long-horizon agentic tasks among open-source models.

MarkTechPost ↗ · February 11 briefing · Open models

ByteDance has introduced Protenix-v1, a comprehensive AlphaFold3 (AF3) reproduction for biomolecular structure prediction, released with code and model parameters under Apache 2.0.

AI Research News ↗ · February 8 briefing · Open models

Waymo is introducing the Waymo World Model, a frontier generative model built on top of Google DeepMind’s Genie 3, to produce photorealistic driving scenes at scale.

AI Research News ↗ · February 8 briefing · Robots and physical AI

Anthropic released Claude Opus 4.6, a major upgrade designed for end-to-end enterprise autonomy. It supports a 1M context window, scored 90.2% on BigLaw Bench, and can autonomously build complex software like a C compiler.

Financial Times / Anthropic ↗ · February 6 briefing

OpenAI launched Frontier, a new platform for managing AI agents like human employees, and GPT-5.3-Codex, which runs 25% faster and extends beyond coding to general professional computer tasks.

OpenAI / TechCrunch ↗ · February 6 briefing · AI that writes code

The Qwen team released Qwen3-Coder-Next, an 80B parameter sparse MoE model designed specifically for coding agents and local development, matching top-tier proprietary performance.

AI Research News ↗ · February 5 briefing · Open models · AI that writes code

The company said that the model was trained on 15 trillion mixed visual and text tokens.

TechCrunch AI ↗ · January 27 briefing · Open models · AI that writes code

Robert Hart /The Verge:Anthropic rolls out a new extension to MCP to let users interact with apps directly inside the Claude chatbot, with support for Asana, Figma, Slack, and others— Anthropic's one step closer to having an everything app. … Anthropic's Claude got a bit livelier today thanks …

TechMeme ↗ · January 26 briefing

Designed to improve robots’ reasoning capabilities, Rho-alpha marks Microsoft’s offering in the growing field of physical AI.

AI Business ↗ · January 26 briefing · Robots and physical AI

NVIDIA Researchers released PersonaPlex-7B-v1, a full duplex speech to speech conversational model that targets natural voice interactions with precise persona control. From ASR→LLM→TTS to a single full duplex model Conventional voice assistants usually run a cascade. Automatic Speech Recognition (ASR) converts speech to text, a language model generates a text answer, and Text to Speech (TTS) converts it back to speech.

AI Research News ↗ · January 19 briefing

Google has been bolstering its AI offerings as it takes on OpenAI, and is now going directly at Apple.

CNBC Technology ↗ · January 14 briefing

AnthropicreleasedCoworkon Monday, a new AI agent capability that extends the power of its wildly successfulClaude Codetool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, largely using Claude Code itself. The launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, positioning Anthropic to compete not just withOpenAIandGooglein conversational AI, but withMicrosoft's Copil...

VentureBeat ↗ · January 12 briefing · AI that writes code

Z.ai:Chinese AI startup Z.ai releases GLM-4.7, an open-weight model that Z.ai says delivers significant improvements in coding performance compared to GLM-4.6— - Core Coding: GLM-4.7 brings clear gains, compared to its predecessor GLM-4.6, in multilingual agentic coding and terminal-based tasks …

TechMeme ↗ · December 22 briefing · Open models · AI that writes code

Mistral AI has released Mistral OCR 3, its latest optical character recognition service that powers the company’s Document AI stack. The model, named as mistral-ocr-2512, is built to extract interleaved text and images from PDFs and other documents while preserving structure, and it does this at an aggressive price of $2 per 1,000 pages with […]The postMistral AI Releases OCR 3: A Smaller Optical Character Recognition (OCR) Model for Structured Document AI at Scaleappeared first onMarkTechPost.

AI Research News ↗ · December 21 briefing

OpenAI releases GPT-5.2-Codex, the most advanced agentic coding model for professional software engineering and defensive cybersecurity, with improvements on long-horizon work through context compaction and stronger performance on large code changes.

OpenAI ↗ · December 18 briefing · AI that writes code

Google is making Gemini 3 Flash the default model in the Gemini app and oAI model for Search

TechCrunch AI ↗ · December 17 briefing

Meta has released SAM Audio, a prompt driven audio separation model that targets a common editing bottleneck, isolating one sound from a real world mix without building a custom model per sound class. Meta released 3 main sizes, sam-audio-small, sam-audio-base, and sam-audio-large. The model is available to download and to try in the Segment Anything […]The postMeta AI Releases SAM Audio: A State-of-the-Art Unified Model that Uses Intuitive and Multimodal Prompts for Audio Separationappeared fir...

AI Research News ↗ · December 17 briefing

The ChatGPT-maker is releasing its "best model yet" as it faces new pressures from Google and other AI competitors.

Wired ↗ · December 11 briefing · AI that writes code

Runway debuts a physics-aware world model that simulates reality to train agents and power video, robotics, and avatar applications.

TechCrunch AI ↗ · December 11 briefing

Google Labs is testing a product that will work with your browser tabs to make web apps for you.

TechCrunch AI ↗ · December 11 briefing

Essential AI, whose CEO co-wrote Google's Attention Is All You Need paper, unveils Rnj-1, an 8B-parameter open model with SWE-bench performance close to GPT-4o. The long-term advancement and equitable diffusion of AI technologies crucially depend on their development in the Open.

TechMeme ↗ · December 7 briefing · Open models

Anthropic says its new Claude Opus 4.5 excels at coding, using computers and assisting users with complex enterprise tasks.

CNBC Technology ↗ · November 24 briefing · Follow the money

Alphabet Inc.’s Google debuted an updated version of its artificial intelligence model, Gemini, that executives said represents a “massive jump” in reasoning and coding ability.

Bloomberg Technology ↗ · November 18 briefing

AI data centers on earth face bottlenecks like intermittent solar power, massive water usage for cooling (up to 1.7 million tons per 40 MW facility over a decade), and grid constraints. Space offers near-constant solar exposure (up to 95% capacity factor in sun-synchronous orbits), zero water needs, and rapid deployment (2-3 months vs. years on ... Read more

Next Big Future AI ↗ · October 28 briefing