Use Opus 5.5 to build a whole film from a long prompt. Use Sonnet 5.5 for small edits, single scenes, scripts and lots of quick tries, at half the price per token. The free plan only has Sonnet, Opus starts with Pro. Before you switch models, try a different effort level first.
Sonnet 5.5 vs Opus 5.5 for motion design comes down to the size of the job. Building a whole film from a long prompt? Use Opus 5.5. Small edits, a single scene or a quick test? Sonnet 5.5 is enough. According to Anthropic, Sonnet is faster and costs half as much per token. Opus is built for long tasks with many steps, and that is exactly what a motion design film is: Claude writes the code for every frame, renders the film, checks the frames and fixes what is off. Every fact in this guide comes from Anthropic, as of November 21, 2026.
The video above is “Raycask”, a launch film of just under 20 seconds in 16:9 for an invented launcher app for the Mac. It has no voice, and every cut sits on a 128 BPM grid. The prompt behind it is about 4,500 words long and sets the colours, fonts, every scene and every sound. A prompt like that is where the model choice starts to matter. If you want a film like this for your own app, copy the prompt from the Raycask page and tell Claude what to swap: name, logo, colour and feature.
Sonnet 5.5 vs Opus 5.5: the facts side by side
Both models read text and images, write code and have the same context window. The difference is price, speed and what Anthropic built each one for. The cards use the numbers from Anthropic's models overview and Claude's pricing page:
Sonnet 5.5
Anthropic calls it “the best combination of speed and intelligence”. Price per million tokens: $2 input, $10 output. Speed: fast. Default effort: high. Context: 1 million tokens. Included in the Free plan.
Opus 5.5
Built, in Anthropic's words, for long-running agentic coding and knowledge work. Price per million tokens: $4 input, $20 output. Speed: moderate. Default effort: medium. Context: 1 million tokens. Included from Pro.
As of November 21, 2026. Prices are API prices. On a subscription you pay a fixed monthly price with usage limits.
Two things stand out. First, Anthropic's own overview says that if you are unsure, start with Opus 5.5 for most workloads. Second, Sonnet's default effort is higher than Opus's. Out of the box, Sonnet thinks harder and Opus is set to be a bit more economical. More on that below.
Why a motion design film is a long task
Claude is not a video model. It doesn't paint frames, it writes code that calculates every frame of the film. A film like Raycask is built in many steps: Claude reads the long prompt, builds scene after scene, renders the film frame by frame to MP4, looks at single frames, finds mistakes and fixes them. It is closer to a small software project than to a chat reply.
That is the kind of work Anthropic's guide to choosing a model lists under Opus 5.5: long, autonomous coding work and vision-heavy workflows where many images get read. Sonnet 5.5 is listed for code generation, content creation, visual understanding and everyday work where speed matters. The start of the Raycask prompt shows how much a model has to keep in mind at once for a film like this:
# Raycask – 19.6-second launch film for a voice dictation feature, 16:9, 60 fps ## 0. What you are building A fast launch film for an invented Mac launcher app called Raycask and its new feature "Dictate to AI". The idea is simple: you think faster than you type, so you should not type to your AI at all. You hold two keys, you talk, and
another 4,473 wordsThis is only the start. Our prompts average around 7,500 words: every scene, every movement, every colour is described precisely. That is why your video comes very close to the example.
Get access nowPick a video
You see exactly what you will get.
Copy the prompt
One click. Structure, style and motion are already in there.
Create your own video
Change topic, style and scenes. No experience needed.
The longer the prompt and the more scenes, beats and timings it holds, the more it matters that the model forgets none of them along the way. How a prompt like this is structured is covered in Claude Motion prompts. How to set up Opus 5.5 for it is in Opus 5.5 motion design prompts.
Which model for which motion design task
In practice it is rarely either/or. A film is made of many small tasks, and not every one needs the bigger model. This overview is our recommendation, based on what Anthropic says about each model:
As of November 21, 2026. DrillUp's recommendation based on Anthropic's model descriptions, not our own benchmark.
A good workflow looks like this: build the film with Opus 5.5. Once it works, switch to Sonnet 5.5 for the many small requests (“make the red darker”, “keep scene two half a second longer”). Anthropic describes this pattern itself: a strong model for the hard decisions, a cheaper one for the bulk of the work. That way most tokens are billed at the lower rate.
Change the effort level before you change the model
Both models have an adjustable effort level. The higher it is, the longer Claude thinks before it answers. That costs time and tokens, but with long prompts it often gives the better result. Anthropic notes that tuning effort is often a better lever than switching models.
This matters for any comparison: Sonnet 5.5 ships on “high”, Opus 5.5 only on “medium”. If you compare both without touching anything, you are comparing two settings as well as two models. For motion design that means:
- Opus drops scenes or shifts timings: set Opus to “high” before you reach for Fable.
- Sonnet takes too long for small edits: set Sonnet lower, for example “medium”. Swapping a word doesn't need deep thinking.
- You want a fair test: give both models the same prompt at the same effort level.
In Claude Code you pick the model with the /model command and the effort level with /effort. Anthropic explains the levels in its effort guide.
What the difference costs
On the API the maths is simple: Opus costs exactly twice as much per token. Here is an example with assumed amounts, not a measurement. Say a film takes 2 million input tokens, because Claude reads the prompt, the code and the frames again and again, and 300,000 output tokens for the code. With Opus 5.5 that is $8 plus $6, so $14. With Sonnet 5.5 it is $4 plus $3, so $7.
The price per token is not the price per film, though. If one model needs two extra rounds before everything is right, the saving is gone. So Sonnet pays off most where the task is clear and rarely needs a second round.
On a subscription you don't pay per token but a fixed price with usage limits. According to the pricing page, how fast you hit your limit also depends on the model you pick. The Free plan only has Sonnet and Haiku. Opus starts with Pro, at $17 a month billed yearly or $20 billed monthly. Claude Motion with its export button is not in any individual plan (Free, Pro, Max, Ultra), only in Team and Enterprise. What to do without those plans is in Claude Motion without a Team plan, and how to turn an animation into an MP4 is in export a Claude animation to MP4.
And when does Fable 5.2 make sense?
Above Opus sits Fable 5.2, since November 9 the most capable model open to all customers. Anthropic recommends it for demanding reasoning and work that runs for hours, or when Opus 5.5 still falls short at a higher effort level. It costs twice as much per token as Opus. Most motion design films don't need it. What Fable can do and which plans include it is covered in Claude Fable 5.2.
Get to a finished film faster
Whichever model you pick, the prompt matters most. A short prompt gets you a draft. A long blueprint with scenes, seconds, colours and sound gets you a film you can post. You don't have to write it yourself. The DrillUp library has ready-made video prompts, each with the film it produced. Take Raycask, for example: copy the prompt, let Opus 5.5 build the film and make the small edits with Sonnet 5.5 afterwards. The DrillUp Academy shows you step by step how to set up Claude for it.
Frequently asked questions
Is Sonnet 5.5 good enough for motion design?
For short scenes, edits to a finished film, scripts and quick tests, yes. For a whole film from a long prompt with many scenes, rendering and checking, Opus 5.5 is the safer choice. Anthropic recommends Opus 5.5 as the starting point for most workloads and lists it for long-running coding work.
How much does Opus 5.5 cost compared to Sonnet 5.5?
On the API, Opus 5.5 costs $4 per million input tokens and $20 per million output tokens. Sonnet 5.5 costs $2 and $10. So Opus is exactly twice the price per token (as of November 21, 2026, Anthropic's models overview).
Can I use Opus 5.5 for free?
No. According to Claude's pricing page, the Free plan includes Sonnet and Haiku but not Opus. Opus starts with Pro, which costs $17 a month billed yearly or $20 billed monthly (as of November 21, 2026).
How do I switch between Sonnet and Opus in Claude Code?
Use the /model command to pick the model and /effort to set the effort level. That way you can build a film with Opus and then make small edits with Sonnet.
Do I need Fable 5.2 for motion design?
Usually not. Anthropic recommends Fable 5.2 for demanding reasoning and work that runs for hours, or when Opus 5.5 still falls short at a higher effort level. Fable costs twice as much per token as Opus.