Progress of Major AI Companies
Following AI related progress of major AI companies (xAI, Anthropic, OpenAI, Meta, for now) based on their official announcements.
OpenAI Releases Agents SDK Evolution
OpenAI released the next evolution of its Agents SDK on April 15, 2026. The update allows developers to create agents that inspect files, run commands, edit code, and tackle long-horizon tasks in secure sandbox environments, with new features like configurable memory and sandbox-aware orchestration. It is generally available via API with standard token-based pricing and initial Python support.
OpenAI Expands Cyber Access with GPT-5.4-Cyber
OpenAI expanded its Trusted Access for Cyber program on April 14, 2026, introducing additional tiers for verified cybersecurity defenders. Highest-tier customers can now request GPT-5.4-Cyber, a fine-tuned GPT-5.4 variant optimized for cyber defense with capabilities like binary reverse engineering. Vetted individuals and enterprises can apply for access through dedicated verification processes.
Meta Releases Muse Spark Model
Meta Superintelligence Labs released Muse Spark, a natively multimodal reasoning model with tool-use, visual chain-of-thought, and multi-agent orchestration capabilities. The model is available in the Meta AI app and via website, with private API preview for select partners. Contemplating mode, which orchestrates parallel-reasoning agents, is rolling out gradually.
Anthropic Launches Managed Agents
Anthropic introduced Managed Agents, a hosted service in the Claude Platform for running long-horizon agents. It decouples agent components like brain, hands, and session for improved reliability, security, and scalability. The service is available now and supports VPCs and custom tools.
OpenAI Launches $100/Month ChatGPT Pro Tier
OpenAI introduced a new $100 per month Pro tier for ChatGPT subscriptions offering 5x more Codex usage than Plus and access to exclusive Pro features including the Pro model and unlimited Instant and Thinking models. The company rebalanced Plus plan Codex limits for steadier weekly usage, ended its Plus promotion, and extended usage promotions for Pro tiers through May 31, 2026. Full pricing details are at https://chatgpt.com/pricing.
xAI Launches Quality Mode for Grok Imagine
xAI has launched Quality mode for Grok Imagine, powered by its most advanced image generation model. This update delivers enhanced details, stronger text rendering in multiple languages, deeper world knowledge, and improved photorealism. The feature is now live on web and mobile apps, with the prior model selectable as Speed mode.
OpenAI Launches ChatGPT in CarPlay
OpenAI has launched ChatGPT in Apple CarPlay, making the voice mode available on-the-go. The feature is rolling out to iPhone users running iOS 26.4 or later in regions where CarPlay is supported.
Meta AI Releases SAM 3.1
Meta AI released SAM 3.1, a drop-in update to SAM 3 introducing object multiplexing for tracking up to 16 objects in a single forward pass. This improves video processing efficiency without accuracy loss, doubling speed to 32 FPS on H100 GPUs for medium-object videos. The model checkpoint and codebase are now openly available.
Anthropic Releases Claude Code Auto Mode
Anthropic released auto mode for Claude Code on March 25, 2026, automating permission decisions with a two-layer safety system to reduce approval fatigue. The feature uses a server-side prompt-injection probe and a Sonnet 4.6-based transcript classifier across three risk tiers. It is available immediately to Claude Code users as a safer alternative to skipping permissions.
OpenAI Launches Product Discovery in ChatGPT
OpenAI announced Product Discovery in ChatGPT on March 24, 2026, launching visually immersive shopping experiences powered by the Agentic Commerce Protocol. Users can now browse products, compare options side-by-side with details like price and reviews, upload images for recommendations, and refine conversationally. The feature is rolling out this week to all ChatGPT plans via web browsers, with mobile apps to follow.
OpenAI Enhances ChatGPT File Management
OpenAI has rolled out enhancements to file management in ChatGPT, including a recent files toolbar, the ability to query uploaded files, and a new Library tab in the web sidebar. The features make it easier to find, reuse, and build upon uploaded and created files. The update is available globally to Plus, Pro, and Business users, with rollout to the EEA, Switzerland, and UK coming soon.
xAI Launches Grok Text-to-Speech API in Beta
xAI announced on March 16, 2026, that Grok's Text to Speech API is now available in beta. The API provides natural speech synthesis in five voices—Eve, Ara, Leo, Rex, Sal—with expressive controls for prosody and emotions, priced at $4.20 per million characters.
OpenAI Releases GPT-5.4 Mini and Nano
OpenAI released GPT-5.4 mini and nano on March 17, 2026. GPT-5.4 mini is available in ChatGPT, Codex, and the API, optimized for coding, computer use, multimodal understanding, and subagents, and runs more than 2x faster than GPT-5 mini. GPT-5.4 nano is available in the API for high-speed, low-cost tasks.
OpenAI Launches Interactive Math/Science in ChatGPT
OpenAI launched interactive learning for math and science in ChatGPT on March 10, 2026. The feature provides an interactive visual module for real-time experimentation with formulas and variables across 70+ topics, including the Pythagorean theorem and ideal gas law. It is rolling out to all logged-in users on consumer and business plans, building on study mode and quizzes.
OpenAI Rolls Out Codex Security Research Preview
OpenAI has rolled out Codex Security, an application security agent that identifies vulnerabilities and proposes fixes, in research preview. It is available to ChatGPT Enterprise, Business, and Edu customers via Codex web with free usage for the next month. Formerly known as Aardvark, it leverages frontier models and the Codex agent, having scanned over 1.2 million commits in beta.
OpenAI Releases GPT-5.4 Thinking and Pro
OpenAI released GPT-5.4 on March 5, 2026, making it available as Thinking and Pro variants in ChatGPT, along with the API and Codex. The model features a 1M token context window in Codex and API, improved factual accuracy and efficiency, and native computer-use capabilities. GPT-5.4 Thinking introduces mid-response steering, rolling out first on Android and web.
OpenAI Releases CoT-Control Evaluation Suite
OpenAI released the open-source CoT-Control evaluation suite and a related research paper on March 5, 2026. The suite comprises over 13,000 tasks assessing reasoning models' ability to control their chain-of-thought to evade monitoring. Frontier models, including GPT-5.4 Thinking, demonstrate low controllability.
OpenAI Releases Learning Outcomes Measurement Suite
OpenAI released the Learning Outcomes Measurement Suite on March 4, 2026, to assess AI's long-term effects on student learning. Developed with academic partners, it features ChatGPT-integrated tools and is undergoing validation with thousands of students.
OpenAI Codex App Now Available on Windows
OpenAI has released the Codex desktop app for Windows users on ChatGPT plans that include Codex access. The app offers a dedicated interface for multitasking with coding agents, previously available only on macOS.
OpenAI Rolls Out GPT-5.3 Instant to ChatGPT
OpenAI has begun rolling out GPT-5.3 Instant to all ChatGPT users and API developers as gpt-5.3-chat-latest. The model offers improved accuracy, fewer refusals and hallucinations, better web search results, and enhanced writing assistance. GPT-5.2 Instant will remain available to paid users until its retirement on June 3, 2026.
Anthropic Launches Claude Code Security Preview
Anthropic has launched Claude Code Security, a new capability integrated into Claude Code on the web. The feature scans codebases for vulnerabilities, suggests targeted patches, and includes verification processes with severity and confidence ratings. It is now available in limited research preview for Enterprise and Team customers, including expedited access for open-source maintainers.
Anthropic Releases Claude Sonnet 4.6
Anthropic released Claude Sonnet 4.6 on February 17, 2026. The model is available via the Anthropic API and demonstrates improved performance over Claude Sonnet 4.5 on benchmarks including SWE-bench Verified and GPQA Diamond. It was deployed under AI Safety Level 3 standards.
OpenAI Adds Lockdown Mode, Risk Labels to ChatGPT
OpenAI introduced Lockdown Mode, an optional advanced security setting for higher-risk ChatGPT users, and Elevated Risk labels for select capabilities in ChatGPT, ChatGPT Atlas, and Codex. These protections aim to defend against prompt injection attacks and provide clearer risk visibility.
OpenAI Releases GPT-5.3-Codex-Spark
OpenAI released GPT-5.3-Codex-Spark, a smaller, text-only model optimized for real-time coding with over 1000 tokens per second and a 128k context window. It is available in research preview for ChatGPT Pro users via the Codex app, CLI, and VS Code extension, running on Cerebras hardware.
OpenAI upgrades ChatGPT deep research to GPT-5.2
OpenAI announced that deep research in ChatGPT is now powered by GPT-5.2, with rollout starting February 10, 2026. New capabilities include connecting to apps and searching specific sites, tracking real-time progress with interruptions, and viewing fullscreen reports.