MaxAI — 릴리스 노트

MaxAI is thrilled to unveil the integration of Claude-Sonnet-4, Anthropic's next-gen hybrid reasoning model with cutting-edge intelligence, excelling in coding, vision, and user-centric tasks.

Key Features

• Hybrid reasoning: Claude-Sonnet-4 merges advanced language understanding, multimodal vision, and tool usage to address complex workflows—enabling effective problem-solving and execution across various tasks.
• Extensive contextual memory: Unlocks a 200K input token window, facilitating in-depth document analysis, large knowledge base Q&A, and smooth handling of multi-part queries.
• Superior coding and automation: Excels in generating, editing, and comprehending code, supporting automated workflows throughout the software development lifecycle.
• Multimodal understanding: Analyzes and interprets charts, graphs, and diagrams. Extracts structured data from visuals—perfect for analytics, research, and reporting.
• Reduced hallucinations and error correction: Capable of self-correction, Claude-Sonnet-4 consistently identifies and rectifies its errors, boosting trust in outputs for knowledge, coding, and visual tasks.
• Extended thinking mode: Emphasizes accuracy and context in response generation, making it suitable for lengthy, complex queries and research scenarios.

We are excited to announce the integration of GPT-Image-1, OpenAI's latest model for image generation and editing, into MaxAI. This powerful new feature enables users to create, modify, and stylize images directly from detailed text prompts, reference images, or both, making it an invaluable tool for design, illustration, marketing, education, and beyond.

Key Features

• Multimodal understanding: Excels at interpreting both text and image references for highly customizable outputs.
• High-quality image creation: Produces diverse, visually stunning results—ideal for art, design, branding, and social tasks.
• Professional-grade editing: Edit, restyle, and combine existing images using plain language instructions or reference media.
• Advanced instruction following: Faithfully captures and executes nuanced, detailed, and complex user prompts.
• Reference-based generation: Guide outputs using additional images to match style, mood, or content.
• Enhanced text rendering: Improved ability to display text inside generated images (placement may still be imperfect in complex layouts).
• Real world knowledge: Uses broad knowledge of objects and scenes for more realistic, context-aware results.
• Content moderation controls: Integrated filtering ensures generated images comply with content guidelines. Moderation strictness can be adjusted as needed.

Instantly transcribe your audio files—just open a file to start live transcription and automatic speaker labeling. Breeze through hour-long lessons in minutes with clickable highlights, allowing you to jump directly to key moments with one click. Experience smarter, faster audio insights with MaxAI.

The Add Source feature in MaxAI has been upgraded for greater flexibility and ease of use. You can now add up to 15 files to any conversation, with each file easily managed and previewed directly from the sidebar.

A new input box experience is also available. With just a click, turn on "Current" to instantly chat with the page you're viewing, or turn on "Search" to retrieve real-time information using AI-powered web search. You can also enable both options to combine insights from the page and the web for the most comprehensive answers.

These enhancements are available to all MaxAI users and are designed to create a more seamless and intuitive user experience.

We are excited to announce that the o4-mini reasoning model is now available in MaxAI. o4-mini delivers state-of-the-art performance in math, coding, and visual reasoning, supporting complex, multi-step tasks across STEM and non-STEM domains.

Key features

• Advanced reasoning in math, coding, and visual tasks, surpassing previous benchmarks
• 93.4% accuracy on AIME 2024 and 92.7% on AIME 2025 math competitions
• 84.3% accuracy on MMMU visual tasks
• 68.1% accuracy (SWE-Bench Verified) for software engineering
• Seamless integration of images and improved contextual conversation
• Enhanced safety with rebuilt refusal and risk detection systems

How to use
o4-mini can be accessed from the AI chat, context menu, AI search, and chat with screenshot features. Each use consumes 1 Pro chat query. Make sure your MaxAI extension is updated for the best experience.

The Edit Question feature is now available in the MaxAI sidebar chat for all users. This update lets you quickly edit and resend your previous messages or adjust prompt variables, giving you greater flexibility and control during every conversation.

Key Details

• Instantly update and resend questions or prompts directly from the sidebar or Web app.
• Enhance workflow efficiency—no need to retype or switch context.
• Easily correct errors or improve your queries as you chat.

We're excited to announce that the GPT-4.1 reasoning model is now available in MaxAI. GPT-4.1 brings state-of-the-art performance in coding, instruction following, vision, and long-context processing to your workflows.

Key features

• Significant boost in coding accuracy (54.6% SWE-bench pass rate) and fewer unnecessary code changes.
• 10.5% improved instruction following; more precise and reliable responses.
• Enhanced understanding of images and long videos—better results on key benchmarks.
• Handles up to 1M tokens for large documents and codebases.
• Knowledge updated to June 2024.

GPT-4.1-mini is now also rolling out in MaxAI. GPT-4.1 mini is a significant leap in small model performance, matching or exceeding GPT-4o in intelligence evaluations while reducing latency by nearly half. Demonstrates remarkable vision capabilities, often outperforming GPT-4o on image benchmarks like MMMU and MathVista.

Key Features

• Advanced Intelligence: Handles complex academic, coding, and vision tasks, matching or exceeding the performance of GPT-4o in intelligence evaluations.
• Ultra-Low Latency: Delivers responses twice as fast as previous models, streamlining real-time applications and long prompts.
• Long Context Support: Supports up to 1 million tokens in a single session, enabling effective analysis of multi-document projects and large codebases.
• Reliable and Precise: Excels at following user instructions literally, ensuring trustworthy results for critical workflows.
• Exceptional Vision Capabilities: Outperforms earlier models in understanding, reasoning, and extracting data from images, charts, and diagrams.

Voice input is now available in MaxAI, allowing you to enter prompts using effortless speech instead of typing. Simply ask your question out loud and get instant answers, saving time and increasing productivity. No more slow typing—just speak and get results.

How it works
To use the microphone, look for the voice input icon in your MaxAI sidebar or Web app. Click to ask your question, and confirm when finished. MaxAI will transcribe your speech and provide a real-time response.

MaxAI now supports Grok-3, xAI's most advanced model yet! It features superior reasoning capabilities refined through large-scale reinforcement learning, enabling it to correct errors, explore alternatives, and deliver accurate answers across mathematics, coding, and knowledge tasks.

Key Features

• Massive 1 million token context window for extensive document processing
• Leading performance in reasoning, mathematical problem solving, and code generation
• Multi-modal capabilities with strong image and video understanding
• Advanced reinforcement learning for complex, step-by-step solutions

Llama-3.3-70B is a high-performance multilingual model that excels in reasoning, coding, and mathematical tasks.

Key Features

• 128K context window and 70B parameter transformer architecture
• Multilingual support—optimized for 8 major languages
• Excels in reasoning, coding, mathematics, and safe, responsible AI responses
• Top performance on coding and academic benchmarks

We're excited to announce a major update to MaxAI's AI reply feature. Our redesigned AI reply is built to enhance your workflow and help you respond to emails, social media, and messaging web apps more efficiently than ever.

What's new in AI Reply?

• Sender's Intent Detection: Instantly understand the sender's intent in every message, giving you vital context for quicker decision making.
• Smart Response Suggestions: Get intelligent, context-aware reply suggestions. Select, edit, and send—no more typing from scratch!
• One-Click Reply Insertion: Easily insert any suggested reply into your draft with a single click.
• Centralized Sidebar Management: Manage all your AI-generated replies from the sidebar for ultimate convenience.

You can now easily add members to your MaxAI team. Please note, each new member requires an additional subscription, and your account will manage their billing. Admins can add or remove team members at any time.

How to Add Members
Go to "Your Plan" and click "People". Select "Add members", enter the new member's email address, and confirm. The new member will join your team instantly.

Pro Chat, a powerful new feature that enhances productivity and workflow with cutting-edge AI capabilities. With Pro Chat, you can choose from a range of advanced AI models—including DeepSeek-R1, o3-mini, GPT-4o, Claude-3.7-Sonnet, Gemini-2.0-Pro, and more—to ensure the best results for any task. Compare and interact with up to six AI models in one conversation for diverse perspectives and deeper insights.

Key Features

• Choose a smarter AI: Select the optimal AI model for your needs
• Ask multiple AI models: Ask and compare answers from up to six AI models
• 5x more web search sources: Access comprehensive, up-to-date information
• In-depth source analysis: Extract richer insights from a broader range of sources
• Advanced reasoning: Solve complex problems with enhanced logical thinking
• More accurate answers: Benefit from higher quality, more reliable responses
• Customizable responses: Personalize language, style, and tone to suit your workflow
• Voice input: Interact hands-free for added convenience and accessibility

Ask AI anything about audio: You can now upload audio files directly, making it easier to work with podcasts, interviews, lectures, and more.

Access Claude-3.7-Sonnet: Anthropic's most advanced hybrid reasoning model, excelling in coding, front-end development, math, and physics. Best for tasks requiring deep reasoning and complex analysis. 200k context window.

Smarter suggested prompts: Save time and complete work faster with more convenient and accurate suggested prompts.

Access Gemini-2.0-Pro: Google's most advanced model for coding and complex reasoning, excelling at tasks requiring deep analysis and world knowledge. Best for developers and complex problem-solving. 2M context window.

Access Gemini-2.0-Flash: Google's latest efficient workhorse model, optimized for high-volume and high-frequency tasks with low latency. Ideal for developers needing fast and reliable performance in scalable applications. 1M context window.

Access o3-mini: OpenAI's latest small reasoning model, optimized for coding, math, and science tasks. Offers low latency and performance advantages. 200k context window.

Access DeepSeek-R1: DeepSeek-AI's advanced RL-based reasoning model, excelling in complex reasoning and achieving performance comparable to leading models in math and code tasks. Best for strategic thinking and analytical tasks.

Access DeepSeek-V3: DeepSeek-AI's powerful MoE model, optimized for efficient inference and large-scale language tasks. Best for extensive data processing and complex analysis.

Create your own tools: You can now create and manage custom prompt templates through the My Tools section, making your workflow more efficient. This feature eliminates the need to repeatedly enter the same prompts, saving you both time and effort.

Access o1: OpenAI's most powerful reasoning model, excelling at advanced reasoning and complex analysis. Best for strategic thinking tasks. 200k context window.

Sidebar rewrite is back: You can now copy and paste text to rewrite it in the sidebar.

The easiest way to view cited source: You can now view cited sources in AI-generated answers effortlessly, without switching tabs or losing focus.

Explore the new design: Enjoy a cleaner, more focused interface that enhances your webpage chat experience. It's simple, friendly, and boosts productivity.

Page summarization now supports switching AI models: You can now select your preferred AI model to summarize and interact with content from webpages, YouTube videos, PDFs, and more.

Access Claude-3.5-Haiku: Anthropic's Claude-3.5-Haiku model is faster, smarter, and better at coding, chat, and data processing. Ideal for apps needing speed and smart thinking.

Access Llama-3.2-90B: Meta's Llama-3.2-90B is a multimodal AI excelling at text and image understanding, including document analysis and visual reasoning.

Use Free AI with vision: You can now take a screenshot or upload any image and ask Free AI to extract text, translate content, solve math problems, explain concepts, and more.

Access o1-preview: Powered by OpenAI, o1-preview thinks deeply about complex problems, excels in science, coding, and math, follows safety guidelines better, and outperforms previous models in tough academic and professional tests.

Access o1-mini: Powered by OpenAI, o1-mini is a faster model that excels at solving math, coding, and science problems quickly and efficiently, with proven success in competitions and top performance in STEM tasks.

Easier and more delightful to use: Experience the new design featuring a cleaner interface. It is now easier to summarize information, answer questions, refine writing, generate drafts, and view AI response details in a more visually appealing format.

MaxAI artifacts: Transform your ideas into tangible content. When enabled, MaxAI will generate visual artifacts like code, documents, and designs in a live preview panel as you chat. The Artifacts feature is available only for GPT-4o and Claude-3.5-Sonnet.

Access FLUX-1-Dev: Powered by Black Forest Labs. FLUX.1 [dev] is Black Forest Labs' most recent open-weight, guidance-distilled model. Derived directly from FLUX.1 [pro], it offers comparable quality and prompt adherence while boasting enhanced efficiency. This model is perfect for developers and researchers looking for high-performance image generation.

Access FLUX-1-Schnell: Powered by Black Forest Labs. Black Forest Labs presents FLUX.1 [schnell], the speediest model in the FLUX.1 family. This variant is optimized for local development and personal use. FLUX.1 [schnell] combines rapid performance with the core strengths of the FLUX.1 series, making it ideal for quick iterations and personal projects.

Access FLUX-1-Pro: Powered by Black Forest Labs. Developed by Black Forest Labs, FLUX.1 [pro] is the pinnacle of image generation AI. This state-of-the-art model excels in prompt adherence, visual quality, and output diversity. With top-tier performance, FLUX.1 [pro] is ideal for professional and enterprise users seeking the best in AI-powered image creation.

Access SD3-Medium: Powered by Stability AI. Developed by Stability AI. Stability AI's most sophisticated open text-to-image model to date. This 2 billion parameter model delivers exceptional image quality, photorealism, and advanced prompt understanding while being resource-efficient for consumer hardware. Featuring improved hands, faces, and typography, SD3 Medium excels in spatial reasoning, compositional elements, and style flexibility. Ideal for both creative professionals and AI enthusiasts seeking high-quality, customizable image generation.

Chat with this image: Interacting with images on any webpage has become incredibly simple. Just hover over an image, click the 'Chat with this image' button, and start asking questions about the image content. Exploring visual information has never been more accessible or convenient.

Access Mistral-Large-2: Next-gen flagship by Mistral AI. 123 billion parameters and 128k context window. Good at code generation, math, and reasoning with enhanced multilingual and 80+ coding languages support.

Access Llama-3.1-405B: Meta's flagship foundation model driving the widest variety of use cases. The world's largest open source AI model with frontier-level capabilities. Expanded context length of 128K tokens for advanced use cases. State-of-the-art performance in general knowledge, steerability, and math. Multilingual support across 8 languages.

Access Llama-3.1-70B: Meta's highly performant model that enables diverse use cases. Longer context length of 128K, state-of-the-art tool use, and overall stronger reasoning capabilities. Ideal for advanced use cases like long-form text summarization, multilingual conversational agents, and coding assistants.

Context menu redesigned: You can now easily switch AI models, change the AI response language, view and copy selected context, all directly from the context menu for improved productivity.

AI rewriter in sidebar: Access your context menu rewrite history right in the sidebar anytime. You can also directly copy and paste text into the sidebar AI rewriter to proofread, improve, and polish any text. It's never been easier.

AI rewriter can be your new context menu in sidebar: You can now choose to always display context menu responses in the sidebar. This way, you can avoid cluttering your browser experience and prevent the context menu from overlapping webpage content.

Access GPT-4o-mini: OpenAI's newest model - Significantly smarter than GPT-3.5-Turbo. As fast as GPT-3.5-Turbo. Supports text and vision capabilities. Improved multilingual understanding.

Resizable context menu: You can now resize the context menu by simply dragging its bottom right corner.

Videos and knowledge pannel in search results: Enjoy deeper insights with Pro Search! Get relevant videos and smart knowledge panels directly in your search results.

Find sources fast with citations: Curious about the sources behind the AI's answers when chatting with web page? Now, you can easily check the citations in each response.

Speak with MaxAI.me to save more time: You can now use your voice to ask questions and prompts to boost your productivity even more.

Access Gemini-1.5-Flash: Gemini 1.5 Flash is Google's lightweight model, optimized for speed and efficiency with multimodal reasoning capabilities.

Images in search results: Enjoy richer answers with Pro Search! Now, you will see relevant images in search results.

Discover with AI search: If you want to catch up on what's happening in the world but aren't sure what to search for, you can now do it with a click.

Never run out of art inspiration: Need inspiration for your next art project? At the start of a new AI art chat, you'll now see examples to spark your creativity.

Access Claude-3.5-Sonnet: The newest model by Anthropic. Delivers better than Opus capabilities and faster than Sonnet speeds. Particularly good at coding, data science, visual processing, agentic tasks, writing, and more.

Effortlessly explore the AI prompt library: A blank page can be intimidating. At the beginning of a new chat, you'll now see examples to help you get started.

Sync your chats: You can now effortlessly switch devices to continue reading, writing, and searching with AI.

Keep exploring in page summary: Summary results now include suggested follow-up questions for any page, PDF, or YouTube video. With just one click, you can easily dive deeper and explore further to get instant answers.

Accept AI-generated reply with 1 click: Speed up your social media responses with 1-click AI-generated replies on LinkedIn, YouTube, and YouTube Studio. Boost your growth!

Pin the context window: Pin the context window to keep generated content open while you browse and interact with the current webpage seamlessly.

Find sources fast with citations: Curious about the sources behind the AI's answers when chatting with PDFs? Now, you can easily check the citations in each response.

Access GPT-4o: OpenAI's new flagship model. Provides GPT-4-level intelligence and with more up-to-date responses. GPT-4o is 2x faster than GPT-4 Turbo. Can reason across audio, vision, and text in real time.

Related questions for AI summary: It's never been easier to dive deeper into your research. With one click on the auto-suggested related questions, you can get instant answers directly on the page, PDF, or video in the Summary tab in the sidebar.

Search with your favorite models: Choose from the latest AI models like GPT-4, Claude 3, and Gemini 1.5 for improved answers and longer context.

Summarize any page in your way: You can now personalize your AI summary by adding your own prompt. This feature allows you to get summaries for any page, PDF, YouTube video, or email. Enjoy the freedom to customize your AI summary experience.

Messenger instant reply: Accelerate your message replies by 10X in 1 click on Messenger.

Enhanced context menu: The context menu is now separate from Chat, giving you a fully immersive experience. You can enjoy the context menu with its AI rewriter and AI reading assistant features.

Telegram instant reply: Accelerate your message replies by 10X with just 1 click on Telegram.

Remaining premium queries: Now you can easily check the queries card in the Settings menu to see the real-time status of your remaining premium queries for this month.

Context window on Microsoft Word: Perfect your writing on Microsoft Word: Improve writing, fix spelling and grammar, make text shorter or longer, simplify language, paraphrase, change tone, and more!

Context window on Google Doc: Perfect your writing on Google Docs: Improve writing, fix spelling and grammar, make text shorter or longer, simplify language, paraphrase, change tone, and more!

Gemini-1.5-pro on MaxAI: Newest and most advanced model by Google. It offers faster speed, 1M token context window, improved modalities, and enhanced safety features.

1-click reply on WhatsApp: Reply to WhatsApp messages with a single click and enjoy the convenience of using your own personalized 1-click reply.

Rename your saved conversations: You can rename your saved conversions in any way you prefer. This makes it easier for you to navigate and find your chat history.

Instant reply on Slack and YouTube Studio: Now, you can use the powerful 1-click reply feature on more messaging web apps such as Slack and YouTube Studio, greatly enhancing your work productivity and taking it to a whole new level.

Your own instant replies everwhere: Create and use your personalized 1-click quick replies for email, social media, and messaging apps, saving you hours every day.

Customize it your way: Whether you want to create your own prompts for quick replies, edit or review drafts, or compose new drafts, we've got you covered! Experience a fully customizable experience in the era of AI!

Instant reply on Discord: Accelerate your message replies by 10X with just 1 click on Discord.

Summarize video: Get the classic video AI summary, including a TL;DR and key takeaways.

Timestamped summary: Navigate YouTube videos effortlessly using summarized paragraphs with timestamped breakdowns.

Summarize comments: Understand the main points of what people think about a YouTube video fast by reading a short AI summary of the top comments.

Show transcript: The fastest and easiest way to review the original transcript of any YouTube video.

Summarize in your preferred method: We are offering more options when summarizing any page, PDF, video, or email - such as TL;DR, Key Takeaways, or both.

Summarize YouTube video comments: Get a summary of the top comments to quickly understand the main opinions and overall sentiment shared by viewers with just one click.

Claude-3-haiku with vision: The fastest model by Anthropic in its intelligence class. Haiku is a versatile solution for a wide range of everyday tasks.

Introducing Pro Art mode: Have full control over art generation without AI-powered art prompts optimization.

Quick reply button enhanced: You will now find the MaxAI quick reply button right next to the default Reply button on email and social media. This means you can reply even faster with one less click.

Gemini-pro with vision: Built from the ground up for multimodality by Google, Gemini on MaxAI now supports image understanding.

Claude-3-opus with vision: The most intelligent model offered by Anthropic, as well as the most capable model in the world on a wide range of benchmarks. It sets a new market standard, offering top-level performance on highly complex tasks, and showing us the frontier of generative AI possibilities.

Claude-3-sonnet with vision: Strikes the ideal balance between intelligence and speed -particularly for high-volume tasks. For the vast majority of workloads, Sonnet is 2x faster than Claude 2 and Claude 2.1 with higher levels of intelligence.

Quoted context: Easily identify the specific context included in each AI prompt you use, and conveniently copy and paste this context for future use whenever needed.

Immersive AI translator: Enable bilingual reading with a single click to improve information retrieval efficiency.

Brand new design: Sleek, redesigned interface—engineered to boost your work productivity to new heights and streamline your workflow effortlessly.

Faster AI model switcher: The newly designed AI model switcher provides a much faster and smoother experience, allowing you to easily switch between various AI models for different purposes.

Ask on any image: Utilize the most advanced GPT-4 Vision model to explain, analyze, and creatively engage with any image you upload, or simply take an AI screenshot.

AI art: 10x your creativity and productivity with the most advanced AI art models. Rapidly turn your visions into stunning images. Simple, powerful, transformative.

Shared Link: Shared link allows you to create and share your chat/search/art conversations with others.

Chat history in sidebar: Easily view and select any previous chat/search/art conversations in the sidebar to conveniently switch between them.