What’s new.

You can now run a workflow once for every file in a folder or every row of a CSV. Add a Bulk node, point it at your product photos or a sheet of prompts, and everything after it runs per row, one at a time or up to three at once. If a row fails the rest still finish and you can re-run only the missing ones, and a new Product mockups in bulk template gets you started.

We added Claude Haiku 5.5, Anthropic's new Haiku model, on OpenRouter and your own Anthropic key. We also added Decider V1.1 27B, Perplexity's decision model, on OpenRouter.

Resources:Claude Haiku 5.5Decider V1.1 27B

More models now work with the key you already have. ElevenLabs v4, v4 Turbo, v3, Flash v2.5, and Scribe v2 now run on OpenRouter, and GPT 6 Luna Decisions runs on your own OpenAI key.

Resources:v4v4 Turbov3Flash v2.5Scribe v2GPT 6 Luna Decisions

We renamed two things to match what these models are now called. The Classify model type is now Decision, and the workflow Select node is now Decide, with D as its shortcut and S moved to Switch. Your saved workflows and model picks carry over as they are.

We added three models. Nano Banana 2.1 is Google's new image model, on Runware, Replicate, OpenRouter, and your own Google key, and Auto now reaches for it in place of Nano Banana 2. Mistral Large 4 joins the text models and GPT 6 Luna Decisions joins the Classify models, both on OpenRouter.

Resources:Nano Banana 2.1Mistral Large 4GPT 6 Luna Decisions

More models now work with the key you already have. Fish Audio S1 and S2 Pro, Gemini 3.8 Flash TTS and Flash-Lite TTS, and Grok Imagine Video 1.5 Lite now run on OpenRouter, and Seedream 5 Flash runs on your own ByteDance key.

Resources:S1S2 ProGemini 3.8 Flash TTSGemini 3.8 Flash-Lite TTSGrok Imagine Video 1.5 LiteSeedream 5 Flash

We added Seed Audio 1.0, ByteDance's audio model, on Runware and OpenRouter. It reads your script aloud and can clone a voice from up to three reference clips you attach. It also makes music and sound effects, so you will find it under Speech, Music, and SFX.

Resources:Seed Audio 1.0

We added Qwen Image 2.1 Pro for images and two Microsoft speech models, MAI-Voice-2.1 and MAI-Voice-2.1 Flash. Qwen Image 2.1 Pro runs on Runware and can edit from up to ten reference images. The two voices run on OpenRouter and read in 23 languages, and the voice you pick sets the language.

Resources:Qwen Image 2.1 ProMAI-Voice-2.1MAI-Voice-2.1 Flash

We added thirteen text models: Claude Sonnet 5.5, GPT 6.1 Sol and Sol Pro, GPT 6 Astra Pro, DeepSeek V4.1 Flash, three GLM-5.3 tiers, four Qwen3.8 tiers, and Muse Spark 1.3. All of them run on OpenRouter, and several also run on Runware or your own vendor key.

Resources:Claude Sonnet 5.5GPT 6.1 SolGPT 6.1 Sol ProGPT 6 Astra ProV4.1 FlashGLM-5.3GLM-5.3 FlashGLM-5.3 PrimeQwen3.8 MaxQwen3.8 Max PrimeQwen3.8 FlashQwen3.8 Omni FlashMuse Spark 1.3

You can now edit a clip from a prompt with FLUX Video Edit and enlarge one with FLUX Video Upscale. Attach a video in the Video workspace and pick either model; the upscaler runs on the clip alone, with no prompt needed.

Resources:Flux Video EditFlux Video Upscale

We added Grok Imagine Video 1.5 Lite, HeyGen Video 1.0, and MiniMax H3 Fast and H3 Max Turbo, four lower-cost models for short clips from text or a starting image. They run on Runware, and two of them also run on OpenRouter or your own xAI key.

Resources:Grok Imagine Video 1.5 LiteVideo 1.0H3 FastH3 Max Turbo

We added two image models: Seedream 5 Flash, a fast, low-cost model for making and editing images, and Ideogram 4.5, which is strong at text inside images. Seedream 5 Flash runs on Runware and OpenRouter, and Ideogram 4.5 on Runware and Replicate.

Resources:Seedream 5 FlashIdeogram 4.5

You can now make full-length songs with Lyria 3.5 using your own Google key. Describe the music, add lyrics or reference images if you like, and CSuite saves the song as an MP3.

Resources:Lyria 3.5

We added Clef and Clef Flash, Cloudflare's open decision models, to Classification. They run on OpenRouter, or on your own computer through Ollama.

Resources:Clef 27BClef Flash 9B

GPT 6 Sol, GPT 6 Luna, Claude Opus 5.5, Claude Fable 5.1, and Grok 4.7 now run on Runware, and Wan 3.0 and Wan 3.0 Prime on Replicate. For transcription, we added Gemini 3.5 Transcribe, and Whisper, both GPT-4o Transcribe tiers, and Fish Audio Transcribe 1 Pro now run on OpenRouter too.

Resources:GPT 6 SolGPT 6 LunaClaude Opus 5.5Claude Fable 5.1Grok 4.7Wan 3.0Wan 3.0 PrimeGemini 3.5 TranscribeWhisperGPT-4o TranscribeGPT-4o Mini TranscribeTranscribe 1 Pro

We added Classification models, small models that read a request and pick from a fixed set of answers. Under Models, choose one or leave it on Auto, and CSuite uses it to understand your requests for Auto model selection, Chat's choice of tool, Text's edit mode, and the workflow Select node. There are seven to start with: Jev, Tev1 4B, Tev1 0.8B, Nimble 9B, Laya, D1, and Solar Decide, on Ollama, OpenRouter, and Runware.

Resources:JevTev1 4BTev1 0.8BNimble 9BLayaD1Solar Decide

You can now use CSuite in 20 languages: English, Spanish, Portuguese, French, German, Italian, Japanese, Korean, Simplified and Traditional Chinese, Russian, Hindi, Arabic, Bengali, Indonesian, Hebrew, Turkish, Vietnamese, Polish, and Ukrainian. Pick yours in Account, Settings, or on the Welcome step of onboarding, and new installs start in your system's language, with Arabic and Hebrew laid out right to left. Chat and Text also understand requests in any language, so asking for an image or a multi-page guide works the same in Spanish or Japanese as in English.

We added FLUX 3 Image, Black Forest Labs' new model for making and editing images, with up to ten reference images and sharp text. It runs on Runware, Replicate, OpenRouter, and your own Black Forest Labs key.

Resources:Flux 3 Image

Exporting a composition is now faster, because CSuite renders each frame at the size you export instead of at your screen's double density. Video inside an export now lands on the exact frame every time, so clips play back smoothly instead of looking choppy.

We added ElevenLabs v4, their most expressive speech model, and v4 Turbo, a faster version at half the price. Both run on your own ElevenLabs key with the voices on your account, and follow audio tags such as [whispers] written into the script.

Resources:v4v4 Turbo

You can now compose longer audio pieces as well as videos, such as a podcast or a narrated story. Open Compositions, which now has its own place on the rail, and write the script line by line, or describe the piece in the AI chat and it writes the lines and adds the speakers for you. Give each speaker a voice and a colour, and each line becomes speech on the timeline beside lanes for music and sound effects, ready to export as MP3, M4A, or WAV.

You can now try an idea cheaply before paying for the full clip. Turn on Draft for Seedance 2.5 or FLUX 3 on Runware to get a quick preview at 480p or 720p, and when you like it, choose Render in 1080p in the Video workspace or next to Extend in a composition. A draft made in a workflow gets the same choice once a File node saves it.

Resources:Seedance 2.5Flux 3 Video

We added Gemini 3.8 Flash, Google's most capable Flash model, to Chat and Text on Runware, OpenRouter, and your own Google key. Gemini 3.8 Flash TTS and Flash-Lite TTS bring Google's new voices to your Google key, with cues such as <sigh> or <laughs>, and Fish Audio Transcribe 1 joins Transcribe through OpenRouter.

Resources:Gemini 3.8 FlashGemini 3.8 Flash TTSGemini 3.8 Flash-Lite TTSFish Audio Transcribe 1

You can now connect Fish Audio with your own key. It runs S2.1 Pro, a free S2.1 Pro tier for testing, S2 Pro and S1 for speech with voice cloning, and Transcribe 1 and Transcribe 1 Pro for transcripts with speaker labels. Find it under the Vendors tab beside the other eleven.

Resources:Fish Audio S2.1 ProS2.1 Pro FreeS2 ProS1Fish Audio Transcribe 1Transcribe 1 ProFish Audio

You can now let a model make a choice inside a workflow. Give a Select node a few inputs and a question, such as "which one is a red fruit?", and it passes on the one that fits. A Preview node shows text, images, video, or audio the moment it is connected, without saving anything to your project.

ByteDance's Seed 2.1 Turbo, Seed 2.0 Lite, and Seed 2.0 Mini now also run through OpenRouter, so you no longer need a ByteDance key to use them.

Resources:Seed 2.1 TurboSeed 2.0 LiteSeed 2.0 Mini

We added Claude Opus 5.5 and Claude Fable 5.1, Anthropic's newest Opus and Fable models, and Grok 4.7, xAI's new flagship model. All three work in Chat and Text through OpenRouter or your own Anthropic or xAI key.

Resources:Claude Opus 5.5Claude Fable 5.1Grok 4.7

We added Muse Voice Transcribe 1.0, Meta's speech-to-text model, to Transcribe through OpenRouter. OpenRouter asks you to confirm your age in its account settings before it runs this model.

Resources:Muse Voice Transcribe 1.0

We added OpenAI's GPT-6 models: Astra, Sol, and Luna, plus Pro versions of Sol and Luna that work longer on hard questions. Astra runs on Runware, OpenRouter, and your own OpenAI key, and in Chat it runs through OpenRouter or OpenAI. Sol and Luna run on OpenRouter and your OpenAI key, and the Pro versions run on OpenRouter.

Resources:GPT 6 AstraGPT 6 SolGPT 6 Sol ProGPT 6 LunaGPT 6 Luna Pro

We added Qwen Image 2.1, Alibaba's new open 7B image model, to the local GGML provider. It writes clean text inside the image, edits images you attach, and runs fully offline once the weights are downloaded.

Resources:Qwen Image 2.1

We added three more vendors you can connect with your own key: Runway, Kling, and ByteDance. Runway runs its Gen-4.5 and Aleph video and Gen-4 Image models, Kling runs its video, image, and audio models, and ByteDance runs Seed 2, Seedream, and Seedance through BytePlus ModelArk. Find them under the Vendors tab beside the other eight.

Resources:RunwayKlingByteDance

You can now ask Chat about things it needs to look up. It searches the web, reads the pages, and lists its sources under the answer. The search runs from your own machine with no extra key, and the Web toggle in the composer switches it off.

Your documents can now have pages. Open the new Pages panel beside Outline to add, drag, and remove them, or ask for "a 5 part guide" and CSuite writes it page by page. Exports now include an index, and you can save a document as an EPUB with one chapter per page.

We have added P-Video-2 Pro, Pruna AI's best video model, on Runware and Replicate. Give it a prompt or a first and last frame and you get a 5 to 15 second clip with sound at up to 768p. Pick Speed when you are trying ideas and Quality for the final take, which costs about twice as much.

Resources:P-Video-2 Pro

You can now write your style or brand rules once instead of repeating them in every prompt. Add them on the new Instructions page under Account and CSuite sends them with everything you generate. Set one for everywhere and one each for Chat, Text, Image, Video, Music, and Sound effects, and a chip in the composer shows when they are on.

You can now turn a recording into text. Choose Transcribe under Audio, drop in an audio or video file, and the transcript appears under the player, where clicking a line jumps to that moment. We added ten models for it, including three Whisper sizes that run on your machine with no key, and you get a subtitle file whenever the model gives timings.

Resources:WhisperGPT-4o TranscribeGPT-4o Mini TranscribeWhisper Large V3Whisper Large V3 TurboMAI-Transcribe 2Scribe v2Whisper BaseWhisper Small

We added six more vendors you can connect with your own key: Google, ElevenLabs, xAI, DeepSeek, MiniMax, and Black Forest Labs. Find them beside OpenAI and Anthropic under the Vendors tab, and add any model from the vendor's own list.

Resources:GoogleElevenLabsxAIDeepSeekMiniMaxBlack Forest Labs

We added two ElevenLabs models. Flash v2.5 is their fastest speech model, available on Replicate or your own ElevenLabs key with the same voices as v3. Sound Effects v2 makes foley, ambiences, and impacts of up to 30 seconds on an ElevenLabs key, and can make them loop seamlessly.

Resources:Flash v2.5Sound Effects v2

You can now connect a model lab directly with your own API key, starting with OpenAI and Anthropic. Add the key in the new Vendors tab under Models and every call goes through that vendor's official SDK. Their text, image, and speech models show up in your regular list, and you can add more from the vendor's own model list.

Resources:OpenAIAnthropic

We added OpenAI's two speech models for anyone using their own OpenAI key, each with ten voices. With Mini TTS you can also write directions for tone, accent, and pacing.

Resources:GPT-4o Mini TTSTTS-1 HD

We gave Chat a big update. You can now see the model's reasoning, drop or paste in several images, follow every step as it works, and have any reply read aloud. If a document is too big for the model, a Digest whole document button lets it read the lot, and each document shows how much the model saw. Chat also names your conversations for you and only makes images or speech when you ask.

We added three Pruna AI models on Runware and Replicate. P-Image makes an image from text fast for about half a cent, P-Image Upscale enlarges an image you attach by up to 16 times, and P-Video-2 makes clips of up to 20 seconds with sound.

Resources:P-ImageP-Image UpscaleP-Video-2

You can now connect OpenRouter. Add your key under Models and one account covers text, chat, images, video, and audio across 61 of our catalog models. Analytics shows you the exact amount OpenRouter billed for each run, not an estimate.

Resources:OpenRouter

We added OpenAI's GPT Audio and GPT Audio Mini on OpenRouter. Both read your text aloud in any of thirteen voices, and Mini costs about a twentieth as much per thousand characters.

Resources:GPT AudioGPT Audio Mini

We added OpenAI's two new image tiers on both Replicate and Runware. Both render text cleanly and edit images you supply, and you can choose from six quality levels, from about a cent an image up to a new max setting.

Resources:GPT Image 2.5 SunburstGPT Image 2.5 Flare

You are no longer limited to our catalog. Under Models you can search a provider's full library, preview a model, set its price, and add it. It joins your regular list with an "Added by you" badge, and your Chat model fills in its type and description for you.

We added Google's Gemini 3.7 Flash to the text models on Runware. It reads a million tokens of context, writes up to 64K, and costs half of what 3.6 Flash does.

Resources:Gemini 3.7 Flash

We fixed text generation and chat stopping after 60 seconds. A run now ends only after five minutes with no output, so your large documents get the time they need.

We added MiniMax H3 Max to the video models on Runware. It is the faster tier of H3, making 5 to 15 second clips with synchronized audio and first and last frame control, at a lower cost per second than H3 itself.

Resources:H3 Max

You can now render MiniMax H3 on Runware at 768p as well as 2K, and in the 21:9 aspect ratio. 768p costs roughly 40% less per second, so we made it the default.

Resources:H3