Claude 3.7 Sonnet for video: retired model and workflow
Quick answer
Claude 3.7 Sonnet was a writing and reasoning model, not a video generator. Anthropic retired its API model on February 19, 2026 and names Claude Sonnet 4.6 as the replacement. Keep this page for the historical release; use an active model for new scripts, shot lists and prompt drafts, then send the approved visual brief to a video generator.
The retirement date and replacement were checked in Anthropic’s current model-deprecation documentation on September 13, 2026.
A red ceramic mug sits on a wooden café table beside a window. Steam rises slowly while the camera makes a gentle push-in. Soft morning light, realistic materials, one continuous shot, no people or readable text.
Turn an approved script into video
Use one short scene or shot as the brief, generate a result, and review the output before building the full sequence.
Try AI Video GeneratorWhat Claude 3.7 Sonnet did in a video workflow
Anthropic announced Claude 3.7 Sonnet in February 2025 as a hybrid reasoning model. Its relevant video-production role was text work: interpreting a brief, drafting a script, breaking narration into scenes, producing shot lists, and refining prompts. It did not render video frames.
Useful input: audience, objective, duration, platform, facts, source material, visual constraints and required call to action.
Useful output: a script with sourced claims, a shot list with one purpose per shot, and generation prompts that separate subject, action, setting, camera and constraints.
Human review: factual accuracy, pacing, rights, brand fit, unsupported claims and whether every proposed shot is feasible.
A current replacement workflow
For new work, select an active Claude model from Anthropic’s model status and deprecation page. Do not silently swap a retired model ID in production; compare the new output format, instruction following, latency, cost and any structured-output assumptions in your own workflow.
After approving the script and shot list, use an actual AI video generator for the visual output. The app or model may impose its own duration, aspect-ratio, reference-image, audio and safety constraints. Adapt each shot rather than asking a language model to promise capabilities the renderer does not provide.
Reusable prompt for a script and shot list
“Using only the supplied sources, write a [duration] video for [audience] that helps them [goal]. Open with the direct answer. For every claim, cite the supplied source or mark it NEEDS SOURCE. Return: narration, on-screen text, shot purpose, visual description, camera instruction, required source asset, and estimated seconds. Preserve these constraints: [list]. Do not claim you generated or tested footage.”
Then review the result with the AI video prompting guide and translate one approved shot at a time into the syntax and limits of the selected renderer.
Frequently asked questions
No. It can produce text, code, scripts, shot lists and prompts. A video model or editing system renders the visual output.
Anthropic lists the API model as retired as of February 19, 2026. Use the current model-status page to choose an active replacement and test the migration.
Yes, if it still calls the retired model ID. Preserve representative inputs and expected output shapes, migrate to an active model, and compare factual accuracy, formatting, latency and cost before releasing the change.
No. It may improve planning or prompt drafting, but rendered quality also depends on the video model, source assets, settings and review process.
Runbo Li is the Co-founder and CEO of Magic Hour, where he builds AI video and image tools for content creation. He is a Y Combinator W24 founder and former Data Scientist at Meta, where he worked on 0-1 consumer social products in New Product Experimentation. He writes about AI video generation, AI image creation, creative workflows, and creator tools.
