Background texture
ModelMax

Supported Models

ModelMax routes requests to the best available provider. All models are accessed through a single API.

Chat models

Text-in, text-out conversation models. Used with POST /v1/chat/completions.

Google Gemini

Model IDInputOutputNotes
gemini-3.1-pro-previewtext, image, audio, videotextLatest Gemini Pro
gemini-3-pro-previewtext, image, audio, videotextGemini 3.0 Pro
gemini-3-flash-previewtext, image, audio, videotextFast, cost-effective
gemini-3.1-flash-image-previewtext, imagetext, imageImage generation capable
gemini-3.1-flash-lite-previewtext, image, audio, videotextLightweight

AWS Bedrock — DeepSeek

Model IDInputOutputNotes
deepseek-r1texttextReasoning model with chain-of-thought
deepseek-v3.1texttextGeneral purpose
deepseek-v3.2texttextLatest general purpose

AWS Bedrock — Qwen

Model IDInputOutputNotes
qwen3-coder-30b-a3btexttextCode-focused, lightweight
qwen3-32btexttextGeneral purpose
qwen3-235b-a22b-2507texttextLarge, capable
qwen3-coder-480b-a35btexttextCode-focused, high capacity
qwen3-next-80b-a3btexttextEfficient MoE architecture
qwen3-vl-235b-a22btext, imagetextVision-language model
qwen3-coder-nexttexttextLatest coder model

AWS Bedrock — MiniMax

Model IDInputOutputNotes
minimax-m2texttextMiniMax M2
minimax-m2.1texttextMiniMax M2.1

AWS Bedrock — Kimi (Moonshot)

Model IDInputOutputNotes
kimi-k2-thinkingtexttextWith reasoning
kimi-k2.5texttextLatest Kimi

Video models

Async video generation. Used with the Queue API (POST /v1/queue/{model}).

Google Gemini — Veo

Model IDSpeedQualityAudio
veo-3.1StandardHighestYes
veo-3.1-fastFastHighYes
veo-3StandardHighYes
veo-3-fastFastGoodYes
veo-2StandardGoodYes

All Veo models support:

  • Text-to-video and image-to-video
  • Up to 8 seconds duration
  • 720p, 1080p, and 4K resolution
  • Audio generation

Capability matrix

CapabilityGemini ChatBedrock ChatQwen VLVeo Video
Text inputYesYesYesYes
Image inputYes—YesYes
Audio inputYes———
Video inputYes———
Text outputYesYesYes—
Image outputPartial———
Video output———Yes
StreamingYesYesYes—

Choosing a model

For general chat: Start with gemini-3-flash-preview (fast, multimodal) or deepseek-v3.2 (strong text reasoning).

For code: Try qwen3-coder-480b-a35b or qwen3-coder-next.

For reasoning: Use deepseek-r1 (returns reasoning_content in the response).

For vision: Use gemini-3-flash-preview (images, audio, video input) or qwen3-vl-235b-a22b (images only).

For video generation: Use veo-3 for quality, veo-3-fast for speed.