Updates

This update includes a major model portfolio refresh, several new model launches, feature adjustments, and bug fixes.

Model Updates

New Model Launches Nano Banana 2.1: Now available, featuring improved visual quality, complex instruction following, text rendering, and consistency across iterative edits. Gemini 4 Argon: Now available. Designed for complex, long-running professional tasks, it supports million-token context and can process text, images, audio, video, and PDFs.

Model Portfolio Updates Claude Series: Claude Sonnet 5.5, Claude Haiku 5.5, and Claude Fable 5.1 are now available. Gemini 3.8 Flash: Now available. Grok 4.7: Now available. DeepSeek V4.1 Flash: Now available with unified model display naming. Seedream 5.0 Flash: Now available, offering faster and more cost-efficient image generation and editing. Seedream 5.0 Pro remains available.

Models Retired Following Same-Series Upgrades Claude Sonnet 4.6 and Claude Sonnet 5 were replaced by Claude Sonnet 5.5. Claude Haiku 4.5 was replaced by Claude Haiku 5.5. Claude Fable 5 was replaced by Claude Fable 5.1. Gemini 3 Flash Preview, Gemini 3.5 Flash, and Gemini 3.7 Flash were replaced by Gemini 3.8 Flash. Grok 4.3 and Grok 4.6 were replaced by Grok 4.7. DeepSeek V4 Flash was replaced by DeepSeek V4.1 Flash. Seedream v4.5 and Seedream 5.0 Lite were replaced by Seedream 5.0 Flash.

Other Models Retired Claude Opus 4.7, GPT-5.5, GPT-5.4 Pro, GPT-5.4, GPT-5.4 Mini, and GPT-4.1 were not retained because multiple newer models in their respective series are already available. GPT OSS 120B, Wan 2.5 Preview, Flux 2, FLUX.1 Kontext Pro, FLUX.1 Kontext Max, Recraft v4, Recraft v3, and Reve were retired based on model stability and usage volume considerations.

Model Fixes

GPT-6 Astra: Fixed compatibility issues affecting certain model parameters.

Feature Adjustments & Bug Fixes

Mobile Web: Fixed an issue that could prevent files from loading under certain conditions. Affiliate Program Page: Corrected the commission information displayed on the Invite Friends – Affiliate Program page.

Google’s next-generation flagship reasoning model, Gemini 4 Argon, is now live on flowith.

Built for your longest and most complex professional work. Whether you’re working through an extensive report, research across multiple sources, or a problem that demands sustained reasoning, Gemini 4 Argon can hold the bigger picture across up to a million tokens of context—connecting what matters and turning scattered information into sharper conclusions.

More than text—it understands your full material. Text, images, audio, video, and PDFs can all become part of its reasoning. From finding signals in charts and cross-checking evidence across sources to turning complex material into clear analysis and plans, Gemini 4 Argon helps you go deeper on the work that matters.

Bring it the hardest questions—the ones that deserve real patience.

Today, we’re rolling out a major model update across text and image creation. Seven new models are now live, so every task can be matched with intelligence that fits.

A refreshed text-model lineup:

Claude Sonnet 5.5 Claude Haiku 5.5 Claude Fable 5.1 Gemini 3.8 Flash Grok 4.7 DeepSeek V4.1 Flash

From demanding research and long-horizon reasoning to fast, high-frequency everyday work, you now have more ways to match each task with the right balance of capability, speed, and efficiency.

A faster new option for image creation.

Seedream 5.0 Flash

Seedream 5.0 Flash is now live, joining Seedream 5.0 Pro in a more focused 5.0 image-model lineup. Whether you’re rapidly exploring an idea or refining visual details through iteration, there’s now a better-fit model for the way you create.

Some previous models will be retired as part of this update. Full replacement details will be shared in the coming changelog. Head to flowith.io and experience the upgrade today!

Google’s newly upgraded image generation and conversational editing model, Nano Banana 2.1, is now live on flowith.

Fast no longer means settling. Building on Flash-level efficiency, Nano Banana 2.1 delivers stronger visual quality, complex prompt adherence, and text rendering—so posters, social visuals, product imagery, and creative concepts can move from idea to a more finished image, faster.

One idea, developed into a consistent world. Across iterative edits and expanding concepts, it preserves characters and visual elements more reliably—so you can refine composition, details, and expression from the first draft instead of starting over every time.

Create faster. Iterate with confidence. Nano Banana 2.1 is ready.

This update includes new model launches, a new video creation feature, and bug fixes.

Model Updates

Claude Opus 5.5: Now available, designed for long-running, complex work including codebase migrations, system audits, performance optimization, and multi-agent collaboration. GPT-6 Sol / Luna: Now available. Sol balances capability, speed, and cost for complex coding, reasoning, and agent workflows, while Luna is optimized for efficient, high-frequency tasks at scale. GPT-6.1 Sol: Now available, offering near-flagship reasoning for long reports, cross-source research, complex data analysis, and multi-step professional tasks.

Features & Optimizations

Code Video: Now available, supporting video generation from text and reference images. AI automatically arranges storyboards, refines visuals, and adds captions, music, and voiceovers. Draft Mode: When users generate videos with Seedance 2.5, they now have the option to enable Draft Mode. Once enabled, the output is fixed at 480p resolution. Users can then choose to regenerate a 1080p version based on a 480p draft they're satisfied with. Follow-up-first Interaction Mode: Fixed an issue where the default model could be displayed incorrectly in certain situations. Knowledge Garden File Uploads: Fixed an issue that could cause upload progress to display incorrectly.

OpenAI’s newly upgraded GPT‑6.1 Sol is now live on flowith.

Near-Astra capability for deeper, more demanding thinking. From long-report analysis and cross-source research to complex data interpretation and proposal writing, GPT‑6.1 Sol can identify what matters across extensive context, connect the logic, and turn scattered material into clear, reliable conclusions.

More than answering a question—reason until the work takes shape. Across multi-step, iterative professional tasks, it stays anchored to the goal and adapts its judgment as new information and requirements emerge—so complex work comes with fewer gaps and less back-and-forth.

Near-flagship reasoning, built for more of your hardest work.

OpenAI’s next-generation models, GPT-6 Sol and GPT-6 Luna, are now live on flowith.

The right intelligence for every pace of work. Sol — powerful reasoning for demanding coding, complex analysis, and agentic workflows, with a strong balance of intelligence, speed, and cost; Luna — efficient intelligence for high-volume, repeatable work at scale.

Built to understand complex context—then execute at the right pace. Bring demanding, multi-step work to Sol, and put fast, frequent tasks in Luna’s hands.

Match every task with exactly the intelligence it needs.

Anthropic’s next-generation flagship model, Claude Opus 5.5, is now live on flowith.

Built for work that is genuinely long and complex. From codebase-wide migrations and system audits to performance optimization and multi-turn agent collaboration, Opus 5.5 keeps hold of the objective over longer runs — understanding the whole system, moving work forward, and checking the details that matter before it is done.

Stronger without slowing down. Compared with Opus 5, Opus 5.5 delivers major gains on complex work while taking fewer steps and using fewer tokens. It does more than write code: it understands the problem, makes the changes, verifies the result, and pushes large tasks toward something truly deliverable.

Bring Claude Opus 5.5 the work that once demanded endless breakdowns and rework.

This update includes new model launches, feature improvements, and a limited-time campaign.

Model Updates

MiniMax H3: Now available, offering multimodal video generation with up to 2K resolution, improved detail preservation, motion-reference support, and synchronized native spatial audio. GPT-6 Astra: Now available for complex end-to-end work, combining deep reasoning, tool use, long-context understanding, and sustained execution across multi-step tasks. GPT Image 2.5: Now available for image generation and editing, featuring Sunburst for precise iterative refinement and Flare for faster, high-quality generation.

Features & Optimizations

Onboarding: Updated the user questionnaire and supporting materials. Video Model Error Messages: Improved error messages for certain video models to provide clearer explanations when generation fails. Referral Commissions: Corrected the commission display on the Invite Friends page to align with the available balance on Impact.

Campaign Launches

GPT-6 Astra Launch Offer: Enjoy limited-time promotional pricing on GPT-6 Astra for two weeks following its launch.

OpenAI’s next-generation image model, GPT Image 2.5, is now live on flowith.

One model family, two creative speeds. Choose Sunburst when precision matters: start from text or a reference image, then keep refining until the result is closer to what you truly imagined. Choose Flare when momentum matters: generate high-quality images faster, so inspiration never has to wait.

Go beyond generating an image and get it right. Whether you are creating from scratch or continuously revising an existing visual, GPT Image 2.5 brings generation and editing into one connected workflow. See an idea sooner, then refine every detail until it feels right.

Think further. Refine deeper. Create faster. GPT Image 2.5 is ready on flowith.

OpenAI’s most powerful model yet, GPT-6 Astra, is now live on flowith.

Built for the hardest end-to-end work. When a task is information-dense, ambiguous, and full of moving parts, GPT-6 Astra does not stop at advice. It understands the objective, breaks down the path forward, uses tools, keeps executing, and adapts as new information emerges, pushing work all the way to a real result.

Deeper reasoning. More sustained judgment. Whether it is mapping an entire project across a million-token context, navigating a complex codebase, researching an unfamiliar domain, analyzing files and the web, or iterating across multiple steps, Astra holds on to what matters and carries the work forward with consistency.

Limited-time launch offer: Get 50% off GPT-6 Astra for the first two weeks.

From “help me think” to “help me get it done.” GPT-6 Astra is ready on flowith.

MiniMax’s next-generation general-purpose multimodal generation model, MiniMax H3, is now live on flowith.

Bring your visual intent to life with greater precision. From complex text and brand marks to product details, interface elements, and carefully defined creative direction, MiniMax H3 understands your prompts and references more deeply, preserving the details that matter in resolutions of up to 2K. Whether you’re creating an ad, product demo, or any video with a clear visual standard, MiniMax H3 helps you stay closer to the original idea.

And it doesn’t just reference a single frame. It understands the full performance. Combine motion and camera references with characters, products, settings, and sound. MiniMax H3 interprets movement, pacing, cinematic language, and atmosphere to create dynamic scenes, complete with native stereo audio generated alongside the visuals.

Bring MiniMax H3 your references, ideas, and details. And turn them into compelling videos.

This update includes new model launches, feature improvements, bug fixes, and a campaign update.

Model Updates

Gemini 3.7 Flash: Now available, supporting fast understanding and analysis of multimodal content, including long documents, images, videos, audio, and PDFs. Wan 3.0: Now available, supporting complete video generation from multimodal materials, including text, images, audio, video, and documents.

Features & Optimizations

Model Output Display: Fixed display issues affecting results from certain models and improved rendering speed. Image Generation Credit Refunds: Fixed an issue that could affect the refund of credits consumed by image-generation requests in certain situations. Video Model Error Messages: Improved error messages to provide more detailed explanations when requests are rejected because of prompts or source materials. Credit Tracking on iOS Canvas: Fixed an issue that could cause incomplete credit-consumption statistics for canvases involving requests from iOS. Follow-up Nodes: Fixed an issue that prevented Follow-up nodes from being read correctly when used as video source materials. Subscription Renewals: Fixed delayed renewal charges affecting certain payment channels, which could cause payment and subscription entitlement timing to fall out of sync. Invite Friends Page: Improved navigation to the referral page. New Onboarding: Updated the onboarding flow.

Campaign Updates

Select Text Model Discounts: The promotional pricing campaign for select text models has ended as scheduled.

Alibaba Cloud’s next-generation all-in-one video model, Wan 3.0, is now live on flowith.

Let your work materials speak for themselves. Product briefs, slides, data charts, web pages, software interfaces and even documents can become the starting point for a video. Wan 3.0 understands both the information and visual structure behind them, turning static content into a complete story.

More than generating a clip, it helps express a complete idea. Bring together text, images, audio, video, and documents to transform product demos, brand concepts, and data stories from source material into work worth watching.

Bring your materials to Wan 3.0 and turn them into stories worth seeing.

Google’s next-generation multimodal reasoning model, Gemini 3.7 Flash, is now live on flowith.

Speed meets depth. From long documents, images, video, audio, and PDFs to complex context, Gemini 3.7 Flash helps you connect the dots and turn multi-source inputs into clear conclusions.

It is built for demanding, high-frequency work. Whether you’re researching, coding, processing data, or planning multi-step tasks, bring deeper thinking to a faster workflow.

Supercharge your everyday workflow.

This update includes new model launches, feature improvements, policy adjustments, bug fixes, and campaign updates.

Model Updates

Seedance 2.5: Now available, supporting longer-form storytelling, multiple reference images, character settings, storyboards, and immersive audiovisual creation. Grok Imagine Image 2.0: Now available for both image generation and editing. Users can upload reference images and use natural-language prompts to adjust content, styles, and visual details. Grok 4.6: Now available, offering enhanced reasoning capabilities for complex questions, multi-step analysis, research, writing, planning, and development.

Features & Optimizations

Nano Banana Models: Improved generation stability for a more reliable creative experience. Video Model Reference Images: Improved the efficiency and reliability of reference-image processing, reducing cases where video models were unable to read reference images. Canvas Node Placement: Fixed an issue that could cause nodes to overlap unexpectedly in certain situations. Seedance Model Controls: Optimized the parameter selection interface for a clearer and smoother experience. Seedance Real-Person Material Review: Optimized the UI of the real-person material review button across the Seedance model series. Its functionality remains unchanged. Shared Canvases: Fixed an issue where deleted canvas content could still be previewed through a share link while logged out. Error Messages: Improved several error messages.

Policy Adjustments

Unlimited Channel Usage: Due to a recent increase in policy-violating usage, we have updated the usage guidelines and safeguards for the Unlimited Channel to protect the experience of users who use their unlimited access appropriately.

Campaign Updates

Daily Check-in Rewards: Updated the daily check-in rewards. The first-day reward now includes Seedance 2.5 video generation trial credits. Select Text Model Discounts: Introduced new promotional pricing for paid users. Free users remain on standard pricing.

xAI’s new model, Grok 4.6, is now live on flowith.

It has stronger reasoning for deeper exploration. From complex questions and multi-step analysis to open-ended creation, Grok 4.6 helps you connect the dots and turn early ideas into actionable conclusions.

You can have smooth experience from inspiration to execution. Whether you’re researching, writing, planning, or iterating through development, bring Grok 4.6 the problems worth exploring beyond the first step.

Start exploring more with Grok 4.6.

xAI’s next-generation image model, Grok Imagine Image 2.0, is now live on flowith.

Turn ideas into visuals with more freedom. From imaginative concepts and nuanced visual styles to detailed compositions and scenes, Grok Imagine Image 2.0 helps bring your descriptions straight into the frame.

Generate and edit in one flow. Upload reference images, then use natural language to refine content, style, or visual details—making every iteration feel more complete.

Bring the image in your mind to Grok Imagine Image 2.0. Start creating.

ByteDance’s next-generation video model, Seedance 2.5, is now live on flowith.

It delivers longer stories with greater control. Whether you start with a single idea or build from multiple references, character concepts, and storyboards, Seedance 2.5 helps turn scattered inspiration into more complete, coherent visual narratives.

Let sound and visuals come together. From atmosphere and action to camera rhythm, go beyond generating a clip and explore more immersive audiovisual storytelling.

Bring your ideas, assets, and imagination to Seedance 2.5 and start creating.

This update includes model updates, feature improvements, service adjustments, and campaign launches.

Model Updates

Kimi K3: Now available, bringing powerful capabilities for complex coding, deep research, data analysis, and knowledge-intensive work. Model Descriptions: Updated across the platform to make it easier to understand each model’s strengths and select the right one for your task.

Features & Optimizations

Text Node Editor: Fixed a display issue that caused complex formulas and functions, including some GPT-5.6 outputs, to appear incorrectly in the editor. Image Preview: Added zoom controls for inspecting image details more easily. Performance & Animations: Improved loading speeds and updated core libraries, including animation and formula libraries, for a smoother experience. Node Toolbar: Added Crop, Upscale, Background Removal, and Vary features to streamline image editing, making it easier for users to reframe images, improves resolution & clarity, isolates the subject, and creates alternative versions for further exploration. Error Messages: Improved error descriptions to help users identify and troubleshoot issues more accurately.

Service Adjustments

Knowledge Garden API: Temporarily taken offline after we identified a significant volume of abusive activity and policy violations.

Campaign Launches

Seedream 5.0 Pro Free Generations Campaign: We launched a limited-time campaign offering 10 free Seedream 5.0 Pro generations. Users can claim their free generations just by logging and clicking. We have also prepared a Seedream 5.0 Pro × Flowith — Create Beyond the Image event on Discord. Join our Discord community and take part in it!

Kimi K3, a powerful model built for complex coding and deep knowledge work, is now live on flowith.

It has deep reasoning for demanding tasks. From navigating multi-module codebases and debugging tricky issues to analyzing long reports, synthesizing information across sources, and researching industries and competitors, Kimi K3 turns scattered inputs into clear conclusions.

And it is more than text understanding. Ask directly about data, charts, screenshots, or product interfaces to quickly uncover insights and define next steps.

Bring Kimi K3 your hardest research, planning, and development work.

We are giving all users a chance to experience the enhanced image quality of Seedream 5.0 Pro. Starting today, every logged-in user could claim 10 free generations specifically for this model.

Zero friction: No credit math required. It simply gives "10 Free Generations". Available immediately: Just select the Seedream 5.0 Pro model to start generating. Earn more: Participate in the [Seedream 5.0 Pro × Flowith: Create Beyond the Image] event on Discord to unlock more credit rewards!

This update includes model updates, feature & optimizations , and campaign launches.

Model Updates

Nano Banana 2 Lite: Now available. Seedream 5.0 Pro: Now available. GPT-5.6 Sol / Terra / Luna: Now available.

Features & Optimizations

Added canvas alignment guides to make moving and aligning nodes easier. Canvas search results are now sorted in reverse chronological order. Optimized the Rerun function: nodes that are generating can no longer be rerun, helping prevent duplicate credit consumption. Optimized node duplication: model parameters are now retained when duplicating nodes.

Campaign Launches

The new user referral program is back online.

The highly anticipated next-generation model family, GPT-5.6, just landed on flowith.

It is the ultimate model matrix for every task. Choose from three tailored tiers:

Sol (flagship reasoning for deep research and complex coding), Terra (the perfect balance for daily workflows), Luna (blazing-fast speed for high-frequency tasks).

With deeper reasoning and sharper analysis, whether you are debugging intricate codebases, parsing long documents, or solving multi-step logic problems, GPT-5.6 delivers unprecedented accuracy and cognitive depth.

Bring it your hardest problems, and choose the exact power you need.

ByteDance's brand new image generation model, Seedream 5.0 Pro, is now live on flowith!

It delivers deep semantic reasoning and versatile formatting. Built for commercial-grade workflows, this model deeply understands complex instructions and physical lighting. It reliably generates accurate multilingual text, complex data charts, and maintains character consistency.

Turn your complex concepts into production-ready graphics, perfectly sized for any channel.

The ultra-fast image generation model, Nano Banana 2 Lite, is officially live on flowith!

It provides lightning-fast generation and ultimate efficiency. Built for rapid prototyping and high-volume visual creation, this lightweight powerhouse delivers stunning images in as little as 4 seconds.

Iterate faster and scale your visual content without the wait.

Anthropic's highly anticipated next-generation model, Claude Sonnet 5, just landed on flowith.

It has unmatched speed and elite reasoning. Setting a new standard for AI handling daily tasks, Sonnet 5 delivers state-of-the-art performance in coding, complex analysis, and content creation with a blazing fast pace.

Supercharge your daily workflow and get more done in less time.

This update includes model adjustments, feature improvements, and campaign launches.

Model Adjustments

Minimax M3: Now officially available. Claude Fable 5: Temporarily removed due to unforeseen third-party limitations (force majeure). Image 1.5: Temporarily taken offline due to upstream service instability.

Features & Optimizations

Media History: Comprehensively upgraded the feature with multiple rounds of performance and UI optimizations. Flowith Mobile Web Version: Launched the mobile browser version, featuring tailored responsive layouts and mobile-optimized interactions. Web Search: Optimized the web search functionality by integrating it natively into supported models. Nano Banana Models Image Downloads: Improved the download logic so users can now directly save the original, uncompressed images. Canvas & Text Generation: Optimized canvas performance for smoother interactions, and enhanced formatting compatibility for diverse output results during text model generation.

Campaign Launches

New User Daily Check-in: This event is now live. Please refer to the event page for detailed rules and rewards.

We've massively expanded Media History in this update.

Flowith is no longer just a temporary workspace. It's designed to turn every generation into a high-value asset for your future projects.

Universal Auto-Archiving: All images, videos, and Agent Neo deliverables (Docs, Webpages, Slides) now consolidate into one central hub. Batch generations automatically group together for a cleaner, distraction-free UI.

Global Timeline Search: Navigate seamlessly with a scalable day/month/year timeline. Stop hunting across canvases, just scroll through time to find exactly what you need, intuitively and fast.

Lossless Context Restoration: Click any past asset to extract its exact prompt and settings, or jump back to the original canvas node in one click for precise refinement.

Seamless Workflow Integration: Access your history directly via the new canvas side drawer. Drag and drop past assets straight into your active canvas to iterate on the fly, keeping your creative momentum alive.

Anthropic's first publicly available Mythos-class model, Claude Fable 5, just landed on flowith.

The most capable model Anthropic has ever public released. State-of-the-art across nearly every benchmark on software engineering, scientific research, vision, and deep knowledge work. And here's what sets it apart: the longer and more complex the task, the larger its lead.

Bring it your hardest problems. That's where it pulls ahead.

——Inspiration doesn't wait for you to get back to your desk

That's why we've completely rebuilt Flowith Mobile Web to be incredibly smooth, fast, and effortless.

We've transferred the power of Flowith into a lightweight, natural chat interface designed perfectly for single-handed use on the go.

Now, you can conveniently capture, chat, and iterate on your ideas with absolute fluidity, wherever you are.

👉 Start flowing on Mobile：http://flowith.io

Anthropic's newest flagship model, Claude Opus 4.8, is now available on flowith!

It provides unmatched reasoning precision and enhanced complex problem-solving ability, which is engineered to tackle your most demanding workflows with absolute accuracy.

As we transition to this more powerful model, Claude Opus 4.6 has been officially retired. Select Opus 4.8 in your workspace now to experience the new standard of performance.

Google's ultra-efficient model, Gemini 3.5 Flash, just landed on flowith!

It has blazing-fast response times and seamless multimodal capabilities. Let it power through your daily tasks and high-frequency workflows with zero friction right now.

OpenAI's latest generation models, GPT 5.5 and GPT 5.5 Pro, are now officially available on flowith.

It has Advanced reasoning and structural precision. Built on a highly optimized architecture, the new models deliver significantly lower hallucination rates and stronger instruction-following capabilities.

Use GPT 5.5 for high-efficiency daily tasks, and deploy GPT 5.5 Pro to tackle your most complex, multi-step workflows with rigorous logic. Experience the new standard of AI productivity today.

Flowith image generation now supports Image Agent. When using GPT Image , 🍌, or other image models, you can choose Agent-powered generation. Start with a simple prompt, and the Agent uses web search to enrich references and guide the visual direction for more stable, complete outputs.

Key improvements:

Web-aware generation: better aligned with recent trends, current topics, and real-world references, with fewer hallucinations

More professional, refined visuals: stronger composition, texture, and detail

Complex images from simple prompts: turn rough ideas into usable visuals faster

Fewer retries: lower prompt-writing effort and a smoother creation process

This is just the beginning for Image Agent. We’ll keep making image creation lighter and faster. Try it out!

The new benchmark in AI image generation, GPT image 2, is now live on flowith!

It is featuring superior multilingual understanding, naturally human-centric aesthetics, and the ability to effortlessly handle complex, high-precision prompts.

SPECIAL OFFER: Enjoy a massive 90% OFF on all GPT image 2 generations for 24 hours! Tap in and unleash your creativity today!