Guides

Models

The model is the AI that plans your lesson, writes the narration and designs the visuals. A stronger model gives clearer explanations and more reliable animations, especially for long or advanced topics.

How it works

  • Leave model out and you get Claude Haiku 5.5 (anthropic/claude-haiku-5.5), which every plan includes.
  • Pass model with any id from the table below to choose one.
  • Each plan includes a set of models that run on our key, at no extra cost.
  • With your own provider key, every model works on every plan, including the ones marked “Your key”.

All models

“Included from” is the cheapest plan that includes the model on our key. Context is how much text the model can consider at once.

ModelIncluded fromContext

Claude Haiku 5.5Fast

anthropic/claude-haiku-5.5

Anthropic · Quick and inexpensive. Clear explanations for focused topics.

Free1M

DeepSeek V4 ProStrong value

deepseek/deepseek-v4-pro-0813

DeepSeek · Excellent at math and step-by-step derivations for its price.

Free1M

Claude Sonnet 5.5Recommended

anthropic/claude-sonnet-5.5

Anthropic · The best all-rounder: well-paced lessons and reliable animations.

Hobby1M

GPT-6.1 SolBalanced

openai/gpt-6.1-sol

OpenAI · OpenAI's flagship. Strong reasoning and tidy, well-structured slides.

Hobby1.1M

Gemini 3.1 ProVisual

google/gemini-3.1-pro-preview

Google · Google's top model. Great with diagrams, charts and science topics.

Hobby1M

Claude Opus 5.5Most capable

anthropic/claude-opus-5.5

Anthropic · Deepest reasoning for long or advanced lectures.

Pro1M

Grok 4.7

x-ai/grok-4.7

xAI · Fast frontier model with a lively teaching voice.

Pro500K

GPT-6.1 Sol Pro

openai/gpt-6.1-sol-pro

OpenAI · Extra reasoning effort on hard technical topics. Slower.

Your key1.1M

Kimi K3

moonshotai/kimi-k3

Moonshot · Strong open model for code-heavy and math lectures.

Your key1M

Qwen 3.8 Max

qwen/qwen3.8-max-0902

Qwen · Alibaba's flagship. Good multilingual explanations.

Your key1M

GLM 5.3

z-ai/glm-5.3

Z.ai · Capable open model at a very low price.

Your key1M

Mistral Large 4

mistralai/mistral-large-4-0

Mistral · European flagship model. Clear, concise narration.

Your key1M

What each plan includes

PlanModels on our key
FreeClaude Haiku 5.5, DeepSeek V4 Pro
HobbyClaude Haiku 5.5, DeepSeek V4 Pro, Claude Sonnet 5.5, GPT-6.1 Sol, Gemini 3.1 Pro
ProClaude Haiku 5.5, DeepSeek V4 Pro, Claude Sonnet 5.5, GPT-6.1 Sol, Gemini 3.1 Pro, Claude Opus 5.5, Grok 4.7
Any plan + your keyEverything above, plus GPT-6.1 Sol Pro, Kimi K3, Qwen 3.8 Max, GLM 5.3, Mistral Large 4.

Your plan’s monthly video limit applies whichever model you use. See Pricing.

Choosing a model

  • Short, focused topics (“What is a hash map?”): Claude Haiku 5.5 is fast and clear.
  • Math and derivations: DeepSeek V4 Pro is excellent for its price; Claude Opus 5.5 and GPT-6.1 Sol Pro handle the hardest proofs.
  • Best all-rounder: Claude Sonnet 5.5. Well-paced lessons and the most reliable animations.
  • Charts, diagrams, science: Gemini 3.1 Pro.
  • Long or advanced lectures: Claude Opus 5.5.
  • Non-English audiences: Qwen 3.8 Max (with your own key).

Using a model

Pass the id as model. If your plan doesn’t include it and you haven’t connected a key that can run it, the request fails with 403 before anything is charged.

request.sh
curl -sS -X POST "$MANIMOTION_API/video/request" \
  -H "Content-Type: application/json" \
  -H "x-api-key: $MANIMOTION_KEY" \
  -d '{
    "prompt": "Prove that there are infinitely many primes",
    "model": "anthropic/claude-sonnet-5.5"
  }'

In the SDKs it’s the same option: videos.generate({ prompt, model: "anthropic/claude-sonnet-5.5" }) in JavaScript, videos.generate(prompt, model="…") in Python.

Listing models in code

GET /models returns the catalog. Send your API key and each model also tells you whether you can use it right now (available) and which key would run it (via: "byok" for your own key, "platform" for your plan). Without a key you get the public list.

models.sh
curl -sS "$MANIMOTION_API/models" \
  -H "x-api-key: $MANIMOTION_KEY"
RESPONSEresponse.json
{
  "default_model": "anthropic/claude-sonnet-5.5",
  "plan": "free",
  "providers": [
    { "id": "openrouter", "name": "OpenRouter", "connected": false },
    { "id": "openai", "name": "OpenAI", "connected": false },
    { "id": "anthropic", "name": "Anthropic", "connected": true },
    { "id": "gemini", "name": "Google Gemini", "connected": false }
  ],
  "models": [
    {
      "id": "anthropic/claude-sonnet-5.5",
      "name": "Claude Sonnet 5.5",
      "vendor": "anthropic",
      "plan": "hobby",
      "byok_only": false,
      "badge": "Recommended",
      "description": "The best all-rounder: well-paced lessons and reliable animations.",
      "context": 1000000,
      "providers": ["anthropic", "openrouter"],
      "available": true,
      "via": "byok"
    },
    {
      "id": "openai/gpt-6.1-sol",
      "name": "GPT-6.1 Sol",
      "vendor": "openai",
      "plan": "hobby",
      "byok_only": false,
      "badge": "Balanced",
      "description": "OpenAI's flagship. Strong reasoning and tidy, well-structured slides.",
      "context": 1050000,
      "providers": ["openai", "openrouter"],
      "available": false,
      "via": null
    }
  ]
}

The SDKs wrap this as models.list().

Any OpenRouter model

With an OpenRouter key connected, you aren’t limited to this list: pass any model id from openrouter.ai/models exactly as OpenRouter writes it.

Retired model ids

Older ids keep working. On our key they run on their replacement; with your own key they’re sent to your provider as written.

Old idNow runs
google/gemini-2.5-flashClaude Haiku 5.5
google/gemini-2.0-flash-001Claude Haiku 5.5
google/gemini-2.0-flashClaude Haiku 5.5
openai/gpt-4o-miniClaude Haiku 5.5
deepseek/deepseek-v3.2DeepSeek V4 Pro
openai/gpt-4oGPT-6.1 Sol
anthropic/claude-3.5-sonnetClaude Sonnet 5.5
anthropic/claude-3-5-sonnetClaude Sonnet 5.5
anthropic/claude-sonnet-4.5Claude Sonnet 5.5
anthropic/claude-opus-4Claude Opus 5.5
anthropic/claude-opus-4.1Claude Opus 5.5