Model APIs

Choose the model.
Preserve the behavior.

Compare the latest supported models first, then keep the application behavior that matters. One call path serves every family, and the differences that matter are preserved.

Compatible mode

Broad tooling support

  • Existing client libraries and agent frameworks work without rewriting the call path.
  • One request shape across families, so a swap does not become a migration.
  • Suited to applications where portability matters more than family-specific behavior.
Native mode

Family-specific fidelity

  • Request and response semantics are preserved as the model family defines them.
  • Reasoning controls, tool-calling conventions and output limits behave as documented upstream.
  • Suited to applications where a family difference is the reason the model was chosen.

Both modes sit on the same account: one set of credentials, one set of spending and network boundaries, one team responsible for the integration. The mode is a per-application decision, not an account-level commitment.

Who it is for

Applications that need choice without losing intent.

Teams can compare model families without assuming that every family behaves the same way.

Compatible

Keep broad client fit.

Use a consistent interaction when the application values portability across selected families.

Native

Keep family semantics.

Use family-specific behavior when request or response forms carry application meaning.

  • Product teamsShipping a feature on a modelOne account and one set of credentials, with the protocol mode chosen per application rather than per company.
  • Platform teamsServing several internal consumersSpending boundaries and network rules per credential, so one team's experiment cannot affect another's budget.
  • Evaluation teamsComparing families before committingBehavior fields are stated per family, so a comparison is not made on a single aggregate score.

What the account receives

Four practical boundaries around model choice.

Model selection

Review families by the work they must perform.

Fit is judged on the work, not on the length of a list.

Behavior choice

Choose compatible or native behavior for each application path.

Family differences stay visible instead of averaging into one score.

Capability review

Confirm tools, visual input, and output form against the selected family.

Assumptions from one model do not follow you into the next.

Ongoing change

Revisit the supported set as model families and workloads change.

Model choice stays an operating decision, not a one-time purchase.

Latest model display

Browse the latest supported models by brand and modality.

The public display presents the latest supported models across language, image, video, embedding and reranking workloads.

Latest supported models

Anthropic
  • Claude Fable 5.1LanguageVisionReasoning
  • Claude Opus 5LanguageVisionReasoning
  • Claude Sonnet 5LanguageVisionReasoning
OpenAI
  • GPT-6 AstraLanguageVisionTools
  • GPT-5.6 SolLanguageVisionTools
  • GPT-Image-2ImageGenerationEditing
  • text-embedding-3-largeEmbeddingMultilingualRetrieval
  • text-embedding-3-smallEmbeddingRetrieval
Google
  • Gemini 3.8 FlashLanguageVisionTools
  • Nano Banana 2ImageGenerationEditing
  • Lyria 3.5AudioMusicGeneration
  • gemini-embedding-2EmbeddingMultimodalRetrieval
  • EmbeddingGemmaEmbeddingMultilingualRetrieval
Moonshot AI
  • Kimi K3LanguageVisionReasoning
  • Kimi K2.6LanguageVisionReasoning
  • Kimi K2.5LanguageVisionReasoning
Z.ai
  • GLM 5.3LanguageToolsReasoning
  • GLM 5.3 FlashLanguageVision
  • GLM 5.2Language
DeepSeek
  • V4 ProLanguageToolsReasoning
  • V4 FlashLanguageToolsReasoning
  • V4 Flash Vision ExpLanguageVisionTools
Qwen
  • Qwen3.8 MaxLanguageVisionTools
  • Qwen3 EmbeddingEmbeddingMultilingualRetrieval
  • Qwen3 RerankerRerankMultilingualRetrieval
  • Qwen-Image Text to ImageImageGeneration
  • Qwen-Image EditImageEditing
MiniMax
  • M3LanguageVisionReasoning
  • H3VideoGenerationEditing
  • Music 3.0AudioMusicGeneration
  • Hailuo 2.3VideoText-to-VideoImage-to-Video
Mistral
  • ShieldstralLanguageVisionSafety
  • Medium 3.5LanguageVisionTools
  • Small 4LanguageReasoningTools
ByteDance Seed
  • Seedream 5.0 ProImageGenerationReasoning
  • Seedance 2.5VideoGenerationEditing
  • Seedance 2.0VideoGenerationEditing
Kling
  • v3.0 4K Text-to-VideoVideoGeneration
  • v3.0 Pro Image-to-VideoVideoGeneration
  • V3.0 Motion ControlVideoEditing
Wan
  • 2.7VideoText-to-VideoImage-to-Video
  • 2.7 Video EditingVideoEditing
Cohere
  • rerank-v4.0-proRerankMultilingualRetrieval
  • rerank-v4.0-fastRerankMultilingualRetrieval

Ready to move a model into production?

Tell us which model families your application depends on, and what has to keep working. A member of the team replies within one business day.

Request access