# Groq Chat Canonical page: https://fuser.studio/models/groq-chat Run OpenAI's open-weight gpt-oss-120b, gpt-oss-20b and gpt-oss-safeguard-20b on Groq in Fuser. Expand prompts, plan shot lists, batch copy variations. ## Specs | Spec | Value | | --- | --- | | Versions | openai/gpt-oss-safeguard-20b, openai/gpt-oss-120b, openai/gpt-oss-20b | | Inputs | Prompt (text), System Prompt (text) | | Output | text | Turn a rough idea into a director's-brief prompt, a numbered shot list or ten caption options before the next node is done waiting Creator: OpenAI [Draft the brief](https://app.fuser.studio) ## How Groq Chat works Write the brief, choose a model tier, and the answer lands fast enough to feed the next node without breaking your flow. ### Write a brief, not a wish Give it a role, a task and an output format, and put the persistent voice and rules in the system prompt. ### Pick the right model Choose the 120b for hard briefs, the 20b for fast bulk variations, or the safeguard model to check text against a written policy. ### Pipe the text onward Send the clean paragraph, shot list or caption set straight into the image or video node waiting downstream. ## What Groq Chat is good at Clean structure, obedient formatting and cinematic vocabulary on request: it's built to turn loose ideas into text other nodes can use. ### Prompt expansion on demand Hand it a six-word concept and get back a cinematographer-style paragraph covering lens, lighting, palette and mood. Name the visual elements you want and it follows the format exactly. ### Numbered shot lists Break a treatment into numbered beats, each with camera move, framing, light and a one-line action. Structured planning is where the reasoning models are strongest. ### Fast bulk variations Tagline sets, caption options and quick rewrites in one pass. Turn on last-message-only so each batch stays independent of the last. ### Notes into strict templates Paste messy moodboard notes and get a tidy art direction brief in the template you specify. The 120b is the pick for multi-constraint briefs; it follows rules more reliably than the lighter model. ## Made with Groq Chat A gallery of briefs and the structured text they return: image-ready paragraphs, shot lists, headline sets and templated briefs. ### Seed concept to image-ready paragraph ### A treatment broken into numbered beats ### Headline variations in one pass ### Scattered notes into a clean brief ### A word-limited script that obeys the count ## What people build with Groq Chat Prompt writers, planners and copy teams reach for it when speed and tidy structure matter more than lyrical flourish. ### Image prompt writers Expand short concepts into detailed, format-clean prompts for the image node next on your canvas. Lock the voice once in the system prompt and every paragraph arrives in the same shape. ### Storyboard and shot-list planners Turn a treatment or ad concept into numbered shots with camera, framing and action. Quick enough to iterate five versions before you open a video node. ### Campaign copy teams Generate dozens of taglines, captions and product lines in a set voice, then tighten the winners. Use the 20b for volume and the 120b when the brief has many rules. ### Art directors writing briefs Reformat moodboard notes, references and client chatter into a strict creative brief template. The safeguard model can then check the result against a written brand policy and return a label with a short rationale. ### Short-form script writers Write tight voiceovers and spot scripts to an exact length and structure. It is a director's-brief writer, so give it the beats and the voice and it delivers clean copy. ## Groq Chat questions ### How fast is Groq Chat for batches of prompt variations? Fast enough to iterate live. Groq lists the default gpt-oss-120b at about 500 tokens per second, so a set of caption or prompt variations typically returns in seconds. Its 131,072-token context window holds a long brief or script excerpt alongside your instructions. Turn on Last Message Only when each variation should start fresh. ### Which model should I use: gpt-oss-120b, gpt-oss-20b or the safeguard model? Use **gpt-oss-120b** (the default) for nuanced briefs, structured planning and strict instruction-following. Use **gpt-oss-20b** for bulk variations, renaming and quick rewrites. Use **gpt-oss-safeguard-20b** only to judge content against a policy you write in the prompt, not as a general creative writer. ### Can Groq Chat look at my image or video? No, it is text-only and cannot see images, video or audio. Attachments should be text such as briefs, scripts or notes. If you need a visual reference described, run it through a vision-capable node first and pass the description in. ### What is it best at for creative work? It is best at expanding a short idea into a detailed image prompt, shot lists from a brief, rapid headline and caption variations, and rewriting copy to a set voice. It favors lists and tidy structure, and it handles lens, lighting and wardrobe vocabulary well when you ask for it. ### Why is my creative writing coming out dry, or refused? It tends toward neutral, slightly stiff prose and can lose the thread on long drafts, so it is a poor fit for voicey long-form fiction. It also refuses or softens many violent, sensitive or style-imitation briefs. Add a voice in the system prompt and keep each call short. ### Who made the models behind it, and when were they released? The models are OpenAI's gpt-oss family, open-weight reasoning models released under Apache 2.0 in August 2025. The gpt-oss-safeguard-20b fine-tune for policy-based classification followed in October 2025. ## Try Groq Chat on Fuser Turn a rough idea into a director's-brief prompt, a numbered shot list or ten caption options before the next node is done waiting [Draft the brief](https://app.fuser.studio) ## Related models - [ChatGPT](https://fuser.studio/models/chatgpt.md) - [OpenRouter Chat](https://fuser.studio/models/openrouter-chat.md) - [Perplexity Chat](https://fuser.studio/models/perplexity-chat.md) - [Google Gemini Chat](https://fuser.studio/models/google-gemini-chat.md) - [Mistral Chat](https://fuser.studio/models/mistral-chat.md) - [Claude Chat](https://fuser.studio/models/claude-chat.md) Node reference: [Groq Chat docs](https://docs.fuser.studio/docs/nodes/llm/groq-chat.md)