Models

[Alibaba logo]

hh1-i2v

Alibaba's HappyHorse 1.0 image-to-video model. Animates a reference image with an optional text prompt. Supports 720P and 1080P output with durations from 3 to 15 seconds.

• Third-party
• Zero data retention

[Alibaba logo]

hh1-t2v

Alibaba's HappyHorse 1.0 text-to-video model. Generates videos from a text prompt with configurable resolution, aspect ratio, and duration (3-15s).

• Third-party
• Zero data retention

[Alibaba logo]

hh1.1-i2v

Alibaba's HappyHorse 1.1 image-to-video model. Animates a reference image with an optional text prompt, with smoother motion, natural skin textures, and improved close-up quality over 1.0. Supports 720P and 1080P output with durations from 3 to 15 seconds.

• Third-party

[Alibaba logo]

hh1.1-r2v

Alibaba's HappyHorse 1.1 reference-to-video model. Takes 1-9 reference images (characters and scenes) and a prompt that choreographs them into a single video, keeping each subject's identity consistent. Supports 720P and 1080P output with durations from 3 to 15 seconds.

• Third-party

[Alibaba logo]

hh1.1-t2v

Alibaba's HappyHorse 1.1 text-to-video model. Generates videos from a text prompt with stronger dynamic expressiveness, better visual quality, and improved instruction following over 1.0. Configurable resolution, aspect ratio, and duration (3-15s).

• Third-party

[Alibaba logo]

qwen-image-3.0-pro

Alibaba's Qwen Image 3.0 Pro generates images from text prompts with a focus on complex layout generation, small-text precision, and multilingual font rendering. Supports up to 6 image variants per call, negative prompts, seed control, and optional prompt rewriting.

• Third-party

[Alibaba logo]

qwen3-max

Alibaba's Qwen 3 Max is a large language model with strong coding, reasoning, and multilingual capabilities, served via DashScope's OpenAI-compatible endpoint.

• Third-party
• Zero data retention

[Alibaba logo]

qwen3.5-397b-a17b

Alibaba's Qwen 3.5 is a 397B-parameter mixture-of-experts model with 17B active parameters, offering strong reasoning capabilities with efficient inference.

• Third-party
• Zero data retention

[Alibaba logo]

qwen3.7-max

Alibaba's Qwen 3.7 Max is the largest and most capable model in the Qwen3.7 series, a next-generation flagship built for the agent-centric era with deep strengths in programming, office and productivity tasks, and long-term autonomous execution, served via DashScope's OpenAI-compatible endpoint.

• Third-party

[Alibaba logo]

qwen3.7-plus

Alibaba's Qwen 3.7 Plus is the cost-effective member of the Qwen3.7 series, pairing strong text capabilities with image and video understanding and full-stack agent-level intelligence for coding, tool use, and GUI-based automation, served via DashScope's OpenAI-compatible endpoint.

• Third-party

[Alibaba logo]

qwen3.8-max

Alibaba's Qwen 3.8 Max is a 2.4-trillion-parameter MoE flagship built for professional-grade coding and long-horizon autonomous work, capable of delivering complete, production-grade projects spanning 10+ days across legal, financial, design, and other specialized domains. Native visual understanding of images and extended video runs through the full plan-execute-verify cycle, served via DashScope's OpenAI-compatible endpoint.

• Third-party

[Alibaba logo]

wan-2.6-image

Alibaba's Wan 2.6 text-to-image model generating images from text prompts with optional negative prompts and customizable dimensions.

• Third-party
• Zero data retention

[Alibaba logo]

wan-2.7-i2v

Alibaba's Wan 2.7 image-to-video model that generates videos from a reference image with optional text prompts. Supports 720P and 1080P output with durations from 2 to 15 seconds.

• Third-party
• Zero data retention

[Anthropic logo]

claude-fable-5

Claude Fable 5 is Anthropic's most capable widely released model, built for the most demanding reasoning and long-horizon agentic work. Adaptive thinking is always on, and the model supports a 1M token context window with up to 128k output tokens per request.

• Third-party

[Anthropic logo]

claude-haiku-4.5

Claude Haiku 4.5 delivers similar levels of coding performance at one-third the cost and more than twice the speed of larger models.

• Third-party
• Zero data retention

[Anthropic logo]

claude-opus-4.5

Claude Opus 4.5 brings further reasoning, coding, and agentic improvements over Opus 4.1, with stronger tool use and tighter instruction following.

• Third-party
• Zero data retention

[Anthropic logo]

claude-opus-4.6

Claude Opus 4.6 is Anthropic's flagship language model built for complex, multi-step work in coding, financial analysis, and legal reasoning. It uses extended thinking to work through complex problems carefully and features a one million token context window.

• Third-party
• Zero data retention

[Anthropic logo]

claude-opus-4.7

Claude Opus 4.7 is Anthropic's most capable generally available model, with a step-change improvement in agentic coding over Claude Opus 4.6. It uses adaptive thinking to calibrate reasoning per task and supports a one million token context window at standard pricing.

• Third-party
• Zero data retention

[Anthropic logo]

claude-opus-4.8

Claude Opus 4.8 is Anthropic's most capable generally available model, with a step-change improvement in agentic coding over Claude Opus 4.7. It uses adaptive thinking to calibrate reasoning per task and supports a one million token context window at standard pricing.

• Third-party
• Zero data retention

[Anthropic logo]

claude-opus-5

Claude Opus 5 is Anthropic's model for complex agentic coding and enterprise work, delivering intelligence close to Claude Fable 5 at half the price. It uses adaptive thinking to calibrate reasoning per task and supports a one million token context window at standard pricing. Unlike Fable 5, Opus 5 has no data retention requirements for general access.

• Third-party

[Anthropic logo]

claude-sonnet-4.5

Claude Sonnet 4.5 is the best coding model to date, with significant improvements across the entire development lifecycle.

• Third-party
• Zero data retention

[Anthropic logo]

claude-sonnet-4.6

Claude Sonnet 4.6 is Anthropic's latest balanced model offering strong coding, reasoning, and agentic capabilities with improved instruction following.

• Third-party
• Zero data retention

[Anthropic logo]

claude-sonnet-5

Claude Sonnet 5 is Anthropic's most agentic Sonnet model yet, built for coding, tool use, reasoning, and long-horizon professional work at lower cost than Opus-class models.

• Third-party

[AssemblyAI logo]

universal-3-pro

AssemblyAI's Universal 3 Pro speech recognition model for high-accuracy transcription.

• Third-party
• Zero data retention

[Black Forest Labs logo]

flux-2-flex

FLUX.2 [flex] is Black Forest Labs' fine-grained control variant of FLUX.2 exposes tunable inference steps, guidance, and prompt upsampling for typography-heavy and production workflows.

• Third-party

[Black Forest Labs logo]

flux-2-max

FLUX.2 [max] is Black Forest Labs' highest-quality image model top editing consistency, strongest prompt following, and grounding search for visualizations of real-time information.

• Third-party

[Black Forest Labs logo]

flux-2-pro-preview

FLUX.2 [pro] Preview is Black Forest Labs' recommended default for production image generation and editing tracks the latest [pro] weights with strong multi-reference support.

• Third-party

[Black Forest Labs logo]

flux-3-video

FLUX 3 Video is Black Forest Labs' video generation model. It generates video from a text prompt (t2v), animates one or more reference images (i2v), or continues an existing clip (v2v), with synchronized audio, up to fhd resolution, and 5-20 second durations.

• Third-party

[ByteDance logo]

seedance-2.0

ByteDance's next-generation video model with a unified multimodal architecture. Generates high-quality video with synchronized audio from text, images, video clips, and audio inputs. Supports multimodal references (up to 9 images, 3 videos, 3 audio files), native audio generation, video editing, video extension, intelligent duration, and adaptive aspect ratio.

• Third-party

[ByteDance logo]

seedance-2.0-fast

Faster variant of ByteDance's Seedance 2.0 video model. Trades some quality for speed while sharing the same multimodal architecture. Supports text-to-video, image-to-video, native audio generation, multimodal references (images, videos, audio), video editing, and video extension.

• Third-party

[ByteDance logo]

seedance-2.0-mini

ByteDance's compact, cost-efficient video generation model from the Seedance 2.0 family. Supports text-to-video, image-to-video, reference video, and reference audio for background music. Ideal for high-volume workloads where speed and cost matter.

• Third-party

[ByteDance logo]

seedance-2.5

ByteDance's next-generation video model with a unified multimodal reference-to-video architecture. Generates video from text, up to 30 reference images, 10 reference videos, and 10 reference audio clips including audio-only input with no image or video required. Supports first/last-frame image-to-video, video editing, video extension, intelligent duration (including automatic selection), and adaptive aspect ratio.

• Third-party

[ByteDance logo]

seedream-4.0

Seedream 4.0 is ByteDance's image creation model that combines text-to-image generation and image editing into a single architecture, offering fast, high-resolution output up to 4K.

• Third-party

[ByteDance logo]

seedream-4.5

Seedream 4.5 builds on 4.0 with multi-reference image support, batch generation, and sequential image generation.

• Third-party

[ByteDance logo]

seedream-5-lite

Seedream 5 Lite is a lighter, faster version of the Seedream 5 family with multi-reference and batch generation support.

• Third-party

[ByteDance logo]

seedream-5-pro

Seedream 5 Pro is ByteDance's high-quality image generation and editing model with text prompts, up to 10 reference images, and 1K, 2K, or explicit pixel-size output controls.

• Third-party

[ElevenLabs logo]

eleven-flash-v2-5

ElevenLabs' low-latency Flash v2.5 text-to-speech model for fast multilingual speech generation.

• Third-party

[ElevenLabs logo]

eleven-multilingual-v2

ElevenLabs' multilingual text-to-speech model for generating natural speech across many languages with ElevenLabs voices.

• Third-party

[ElevenLabs logo]

eleven-turbo-v2-5

ElevenLabs' Turbo v2.5 text-to-speech model balancing high-quality voice generation with low latency across 32 languages.

• Third-party

[ElevenLabs logo]

eleven-v3

ElevenLabs' latest text-to-speech model for highly expressive, natural speech generation with advanced voice control.

• Third-party

[ElevenLabs logo]

music-v2

ElevenLabs Music v2 composes songs and instrumental tracks from a prompt or detailed composition plan.

• Third-party

[Google logo]

gemini-2.5-flash

Google's fast multimodal Gemini 2.5 model with strong reasoning and a 1M token context window.

• Third-party
• Zero data retention

[Google logo]

gemini-2.5-flash-lite

Google's lightest and most cost-efficient Gemini 2.5 model for high-throughput tasks.

• Third-party
• Zero data retention

[Google logo]

gemini-2.5-pro

Google's most capable Gemini 2.5 model with strong reasoning, thinking support, and a 1M token context window.

• Third-party
• Zero data retention

[Google logo]

gemini-3-flash

Gemini 3 Flash is Google's fast multimodal model with frontier intelligence, superior search, and grounding capabilities.

• Third-party
• Zero data retention

[Google logo]

gemini-3.1-flash-lite

Google's lightest and most cost-efficient Gemini model for high-throughput tasks.

• Third-party
• Zero data retention

[Google logo]

gemini-3.1-flash-tts

• Third-party
• Zero data retention

[Google logo]

gemini-3.1-pro

Google's most intelligent Gemini model with improved reasoning, a medium thinking level, and a 1M token context window.

• Third-party
• Zero data retention

[Google logo]

gemini-3.5-flash

Gemini 3.5 Flash is Google's fast multimodal model with frontier intelligence, superior search, and grounding capabilities.

• Third-party

[Google logo]

gemini-3.5-flash-lite

Gemini 3.5 Flash-Lite is a low-latency, cost-effective multimodal model optimized for high-throughput, low-cost execution for subagent tasks and document parsing.

• Third-party

[Google logo]

gemini-3.6-flash

Gemini 3.6 Flash provides sustained frontier-level intelligence optimized for real-world tasks at a higher speed and lower cost, excelling at code generation, agentic execution, and spatial reasoning.

• Third-party

[Google logo]

gemini-3.7-flash

Gemini 3.7 Flash is a highly capable, natively multimodal reasoning model optimized for agentic workflows and real-world tasks.

• Third-party

[Google logo]

imagen-4

Google's latest image generation model producing high-quality, photorealistic images from text prompts with support for multiple aspect ratios.

• Third-party
• Zero data retention

[Google logo]

nano-banana

Google's fast image generation model producing high-quality images from text prompts.

• Third-party
• Zero data retention

[Google logo]

nano-banana-2

Google's second-generation image generation model with improved quality and speed.

• Third-party
• Zero data retention

[Google logo]

nano-banana-2-lite

Google's fastest Gemini image generation model for rapid image creation and iteration.

• Third-party
• Zero data retention

[Google logo]

nano-banana-pro

Google's higher-quality image generation model with improved detail and prompt adherence.

• Third-party
• Zero data retention

[Google logo]

veo-3.1

Google's latest video generation model with improved quality, motion, and audio generation.

• Third-party
• Zero data retention

[Google logo]

veo-3.1-fast

A faster version of Veo 3.1 optimized for lower latency while maintaining high-quality video and audio output.

• Third-party
• Zero data retention

[Inworld logo]

tts-1.5-max

Highest-quality text-to-speech with under 200ms latency, emotion control, and 15-language support.

• Third-party
• Zero data retention

[Inworld logo]

tts-1.5-mini

Ultra-fast, cost-efficient text-to-speech with approximately 120ms latency and 15-language support.

• Third-party
• Zero data retention

[Inworld logo]

tts-2

Inworld's most powerful and expressive text-to-speech model. Builds on TTS 1.5 with rich expressive speech, real-time latency, natural language steering (e.g. [whisper], [say excitedly]), and stronger multilingual support across 15 production languages plus 90+ experimental languages.

• Third-party
• Zero data retention

[MiniMax logo]

hailuo-2.3

A high-fidelity video generation model optimized for realistic human motion, cinematic VFX, expressive characters, and strong prompt and style adherence across text-to-video and image-to-video workflows.

• Third-party
• Zero data retention

[MiniMax logo]

hailuo-2.3-fast

A lower-latency version of Hailuo 2.3 that preserves core motion quality, visual consistency, and stylization while enabling faster iteration.

• Third-party
• Zero data retention

[MiniMax logo]

m2.7

MiniMax's M2.7 language model with multilingual capabilities.

• Third-party
• Zero data retention

[MiniMax logo]

m3

MiniMax's M3 language model with frontier coding and agentic capabilities, a 1M token context window, and multilingual support.

• Third-party
• Zero data retention

[MiniMax logo]

music-2.6

MiniMax's music generation model that creates full-length songs with vocals from text prompts and lyrics, or instrumental tracks. Supports BPM/key control and auto-generated lyrics.

• Third-party
• Zero data retention

[MiniMax logo]

speech-2.8-hd

MiniMax Speech 2.8 HD focuses on studio-grade audio generation with emotion control, multilingual support (40+ languages), and voice cloning.

• Third-party
• Zero data retention

[MiniMax logo]

speech-2.8-turbo

MiniMax Speech 2.8 Turbo turns text into natural, expressive speech with voice cloning, emotion control, and 40+ language support at faster speeds.

• Third-party
• Zero data retention

[Moonshot AI logo]

kimi-k3

Kimi K3 is Moonshot's flagship 2.8 trillion-parameter model, built on Kimi Delta Attention (a hybrid linear attention mechanism) with Attention Residuals. It offers native visual understanding, always-on reasoning, and a 1M-token context window for long-horizon coding, knowledge work, and deep reasoning tasks.

• Third-party

[OpenAI logo]

gpt-4.1

OpenAI's flagship GPT model for complex tasks with a million-token context window.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-4.1-mini

Fast, affordable version of GPT-4.1 with a million-token context window.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-4.1-nano

GPT-4.1 Nano is OpenAIs smallest and cheapest GPT-4.1 variant, optimized for high-throughput, low-latency tasks.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-4o

GPT-4o is OpenAIs multimodal flagship, accepting text and images and responding quickly across a wide range of tasks.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-4o-mini

GPT-4o Mini is the lightweight, low-cost variant of GPT-4o, well suited to high-volume tasks with multimodal inputs.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-4o-transcribe

A speech-to-text model that uses GPT-4o to transcribe audio with improved word error rate and better language recognition compared to original Whisper models.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5

OpenAI's model excelling at coding, writing, and reasoning.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5-mini

GPT-5 Mini is the lightweight, low-cost variant of GPT-5, well suited to high-volume coding and reasoning tasks.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5-nano

GPT-5 Nano is OpenAIs smallest GPT-5 variant, optimized for low latency and cheap, high-throughput tasks.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.1

GPT-5.1 is OpenAIs incremental improvement over GPT-5, with stronger coding, reasoning, and writing.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.4

GPT-5.4 is OpenAI's flagship model with strong coding, reasoning, and multimodal capabilities.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.4-mini

GPT-5.4 mini is a smaller, faster, and more cost-efficient version of GPT-5.4 for lightweight tasks.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.4-nano

GPT-5.4 nano is OpenAI's smallest and fastest model, optimized for edge and low-latency use cases.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.4-pro

GPT-5.4 pro uses OpenAI's Responses API with built-in tools, improved reasoning, and stateful context management.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.5

GPT-5.5 is OpenAI's flagship model with strong coding, reasoning, and multimodal capabilities.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.5-pro

GPT-5.5 pro uses OpenAI's Responses API with built-in tools, improved reasoning, and stateful context management.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-5.6-luna

GPT-5.6 Luna is an OpenAI GPT-5.6 model optimized for cost-sensitive workloads, using the Responses API for efficient text generation.

• Third-party

[OpenAI logo]

gpt-5.6-sol

GPT-5.6 Sol is OpenAI's frontier GPT-5.6 model for complex professional work, using the Responses API for reasoning and stateful context management.

• Third-party

[OpenAI logo]

gpt-5.6-terra

GPT-5.6 Terra is an OpenAI GPT-5.6 model that balances intelligence and cost, using the Responses API for reasoning and stateful context management.

• Third-party

[OpenAI logo]

gpt-image-1.5

OpenAI's image generation model that creates and edits images from text prompts, supporting multiple quality levels and output sizes.

• Third-party
• Zero data retention

[OpenAI logo]

gpt-image-2

OpenAI's next-generation image model that creates and edits images from text prompts, with support for multiple quality levels, sizes, and output formats. Note: transparent backgrounds are not supported use openai/gpt-image-1.5 for transparent PNGs.

• Third-party
• Zero data retention

[OpenAI logo]

o3

o3 is OpenAIs general-purpose reasoning model, balancing strong analytical performance with reasonable latency and cost.

• Third-party
• Zero data retention

[OpenAI logo]

o3-mini

o3-mini is the lightweight, low-cost reasoning variant of o3, well suited to quick analytical tasks at scale.

• Third-party
• Zero data retention

[OpenAI logo]

o4-mini

OpenAI's fast, lightweight reasoning model optimized for multi-step problem solving at lower cost.

• Third-party
• Zero data retention

[OpenAI logo]

tts-1

OpenAI's text-to-speech model optimized for real-time use with low latency.

• Third-party
• Zero data retention

[OpenAI logo]

tts-1-hd

OpenAI's high-definition text-to-speech model producing higher quality audio output.

• Third-party
• Zero data retention

[PixVerse logo]

v5.6

Pixverse v5.6 is a video generation model supporting text-to-video and image-to-video with audio generation, customizable aspect ratios, and up to 1080p output.

• Third-party
• Zero data retention

[PixVerse logo]

v6

Pixverse v6 is the latest Pixverse video model with support for up to 15-second videos, customizable duration from 1 to 15 seconds, and audio generation.

• Third-party
• Zero data retention

[Pruna AI logo]

p-image

Pruna's P-Image is an ultra-fast text-to-image model with automatic prompt enhancement and 2-stage refinement, combining exceptional speed with high-quality output and flexible aspect ratios.

• Third-party

[Pruna AI logo]

p-image-edit

Pruna's P-Image-Edit edits and composes 1-5 reference images with text instructions. It supports complex compositions, style transfers, and targeted edits with flexible output aspect ratios.

• Third-party

[Pruna AI logo]

p-image-try-on

Pruna's P-Image Try-On virtually fits one or more garments onto a person's photo. Provide a photo of a person plus garment reference images and the model realistically dresses the person in the provided garments.

• Third-party

[Pruna AI logo]

p-image-upscale

Pruna's P-Image-Upscale increases image resolution using AI, targeting 1-128 megapixels with optional detail and realism enhancement for sharper, cleaner results.

• Third-party

[Pruna AI logo]

p-video

Pruna's P-Video is a premium video generation model supporting text-to-video, image-to-video, and audio-conditioned generation up to 1080p at 24 or 48 fps, with configurable duration up to 20 seconds.

• Third-party

[Pruna AI logo]

p-video-animate

Pruna's P-Video-Animate takes a source video and a subject reference image, then animates the referenced subject using the motion and audio from the source video.

• Third-party

[Pruna AI logo]

p-video-avatar

Pruna's P-Video-Avatar generates talking-head videos from a single portrait image driven by a text script or audio file, with multiple voices, languages, and output resolutions.

• Third-party

[Pruna AI logo]

p-video-replace

Pruna's P-Video-Replace takes a source video and one or more identity reference images, then places the referenced person or people into the video while preserving the source motion and audio.

• Third-party

[Recraft logo]

recraftv3

Recraft V3 is the previous-generation text-to-image model from Recraft, well-suited to design-quality compositions, brand-aware imagery, and accurate text rendering.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4

Recraft V4 generates art-directed images with strong composition, accurate text rendering, and design taste built in. Fast and cost-efficient at standard resolution.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1

Recraft V4.1 generates art-directed images tuned for high aesthetics, with strong composition, accurate text rendering, and refined design taste. Fast and cost-efficient at standard resolution.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1-pro

Recraft V4.1 Pro generates high-resolution, art-directed images at 2048px+ tuned for high aesthetics, with strong composition, text rendering, and refined design taste. Built for print and production work.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1-pro-vector

Generate detailed, high-resolution SVG vector graphics from text prompts with high aesthetic quality, fine geometry, scalable to any size for print and design work.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1-utility

Recraft V4.1 Utility is a general-purpose text-to-image model balancing quality and flexibility for a wide range of everyday use cases at standard resolution.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1-utility-pro

Recraft V4.1 Utility Pro is a general-purpose text-to-image model producing high-resolution 2048px+ output for a wide range of production and print use cases.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1-utility-pro-vector

Generate detailed, high-resolution SVG vector graphics from text prompts with a general-purpose model, scalable to any size for print and large-scale design work.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1-utility-vector

Generate production-ready SVG vector graphics from text prompts with a general-purpose model suited for a wide range of design and illustration tasks.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-1-vector

Generate production-ready SVG vector graphics from text prompts with high aesthetic quality, clean geometry, structured layers, and editable paths.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-pro

Recraft V4 Pro generates high-resolution, art-directed images at 2048px+ with strong composition, text rendering, and design taste. Built for print and production work.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-pro-vector

Generate detailed, production-ready SVG vector graphics from text prompts with fine geometry, scalable to any size for print and design work.

• Third-party
• Zero data retention

[Recraft logo]

recraftv4-vector

Generate production-ready SVG vector graphics from text prompts with clean geometry, structured layers, and editable paths.

• Third-party
• Zero data retention

[RunwayML logo]

aleph-2

RunwayML's video editing model. Edit one frame to update your whole video, make changes across multiple shots, and work with up to 30 seconds of video. Supports keyframe-guided editing for precise control over specific moments in the clip.

• Third-party

[RunwayML logo]

gen-4.5

RunwayML's video generation model supporting both text-to-video and image-to-video with customizable duration, aspect ratio, and content moderation controls.

• Third-party

[Vidu logo]

q3-pro

Vidu Q3 Pro is a high-quality video generation model supporting text-to-video, image-to-video, and start/end-frame-to-video workflows with audio and up to 16-second clips.

• Third-party
• Zero data retention

[Vidu logo]

q3-turbo

Vidu Q3 Turbo is a faster version of Vidu Q3 optimized for lower latency video generation while maintaining audio support and up to 16-second clips.

• Third-party
• Zero data retention

[xAI logo]

grok-4.20-0309-non-reasoning

xAI's Grok 4.20 non-reasoning model. Skips the thinking trace for fast, single-pass responses while keeping the same training as the reasoning variant.

• Third-party
• Zero data retention

[xAI logo]

grok-4.20-0309-reasoning

xAI's Grok 4.20 reasoning model. Uses extended thinking to work through complex problems, returning a reasoning trace alongside the final answer.

• Third-party
• Zero data retention

[xAI logo]

grok-4.20-multi-agent-0309

xAI's Grok 4.20 multi-agent model with a 2M-token context window. Multiple agents collaborate in parallel to perform deep research tasks, with function calling, structured outputs, and reasoning capabilities.

• Third-party
• Zero data retention

[xAI logo]

grok-4.3

xAI's Grok 4.3 model with a 1M-token context window and strong agentic tool calling with minimal hallucinations. Accepts text and image inputs, and supports function calling, structured outputs, and configurable reasoning effort (none, low, medium, high).

• Third-party
• Zero data retention

[xAI logo]

grok-4.5

xAI's Grok 4.5, a frontier model built for coding, agentic tasks, and knowledge work. Accepts text and image inputs, and supports function calling, structured outputs, and configurable reasoning effort (low, medium, high).

• Third-party
• Zero data retention

[xAI logo]

grok-4.6

xAI's Grok 4.6, a flagship reasoning model for coding, agentic tasks, and visual work. Accepts text and image inputs, and supports function calling and structured outputs.

• Third-party
• Zero data retention

[xAI logo]

grok-imagine-image

xAI's Grok Imagine image model. Generates and edits images from text and reference-image inputs with configurable aspect ratio and resolution.

• Third-party
• Zero data retention

[xAI logo]

grok-imagine-image-2.0

xAI's Grok Imagine Image 2.0 is a precise image generation and editing model for creative work, with strong instruction following, typography, layout, and reference-image preservation.

• Third-party

[xAI logo]

grok-imagine-image-quality

xAI's higher-fidelity text-to-image model optimized for sharper details, more accurate compositions, and stronger text rendering. Supports image editing via reference images and masks. Trades speed for quality compared to grok-imagine-image. Default output at 2k resolution.

• Third-party
• Zero data retention

[xAI logo]

grok-imagine-video

xAI's video generation model. Generates, edits, and extends videos from text and image inputs with native synchronized audio including dialogue, sound effects, and music. Supports multiple creative modes (normal, fun, custom).

• Third-party
• Zero data retention

[xAI logo]

grok-imagine-video-1.5-preview

xAI's next-generation video generation model. Generates, edits, and extends videos from text and image inputs. Supports multiple aspect ratios and resolutions with improved quality over the previous generation.

• Third-party
• Zero data retention

[xAI logo]

grok-stt

xAI's Grok speech-to-text model. Transcribes audio files into text across 25 languages with word-level timestamps, multichannel transcription, speaker diarization, and key-term biasing.

• Third-party
• Zero data retention

[xAI logo]

grok-tts

xAI's Grok text-to-speech model. Generates high-fidelity spoken audio in 5 expressive voices (eve, ara, rex, sal, leo) with 20+ supported languages. Supports inline speech tags for laughter, whispers, and pauses.

• Third-party
• Zero data retention

[xAI logo]

grok-voice

xAI's real-time voice conversation model with low-latency audio input and output streaming.

• Third-party
• Zero data retention

[Deepgram logo]

aura-1

Aura is a context-aware text-to-speech (TTS) model that applies natural pacing, expressiveness, and fillers based on the context of the provided text. The quality of your text input directly impacts the naturalness of the audio output.

• Cloudflare-hosted
• Batch
• Partner
• Real-time

[Deepgram logo]

aura-2-en

Aura-2 is a context-aware text-to-speech (TTS) model that applies natural pacing, expressiveness, and fillers based on the context of the provided text. The quality of your text input directly impacts the naturalness of the audio output.

• Cloudflare-hosted
• Batch
• Partner
• Real-time

[Deepgram logo]

aura-2-es

Aura-2 is a context-aware text-to-speech (TTS) model that applies natural pacing, expressiveness, and fillers based on the context of the provided text. The quality of your text input directly impacts the naturalness of the audio output.

• Cloudflare-hosted
• Batch
• Partner
• Real-time

[Meta logo]

bart-large-cnn

Beta

BART is a transformer encoder-encoder (seq2seq) model with a bidirectional (BERT-like) encoder and an autoregressive (GPT-like) decoder. You can use this model for text summarization.

• Cloudflare-hosted
• Deprecated

[BAAI logo]

bge-base-en-v1.5

BAAI general embedding (Base) model that transforms any given text into a 768-dimensional vector

• Cloudflare-hosted
• Batch

[BAAI logo]

bge-large-en-v1.5

BAAI general embedding (Large) model that transforms any given text into a 1024-dimensional vector

• Cloudflare-hosted
• Batch

[BAAI logo]

bge-m3

Multi-Functionality, Multi-Linguality, and Multi-Granularity embeddings model.

• Cloudflare-hosted

[BAAI logo]

bge-reranker-base

Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. You can get a relevance score by inputting query and passage to the reranker. And the score can be mapped to a float value in [0,1] by sigmoid function.

• Cloudflare-hosted

[BAAI logo]

bge-small-en-v1.5

BAAI general embedding (Small) model that transforms any given text into a 384-dimensional vector

• Cloudflare-hosted
• Batch

[DeepSeek logo]

deepseek-r1-distill-qwen-32b

DeepSeek-R1-Distill-Qwen-32B is a model distilled from DeepSeek-R1 based on Qwen2.5. It outperforms OpenAI-o1-mini across various benchmarks, achieving new state-of-the-art results for dense models.

• Cloudflare-hosted
• Reasoning

[DeepSeek logo]

deepseek-v4-flash-0731

DeepSeek-V4-Flash-0731 is the official release of DeepSeek-V4-Flash, superseding the preview version, with substantially enhanced agentic capabilities.

• Cloudflare-hosted
• Function calling
• Reasoning

[DeepSeek logo]

deepseek-v4-pro-0813

DeepSeek V4 Pro is a high-capability reasoning model from DeepSeek with a one million token context window, built for long-horizon agentic workflows and complex, multi-step problem-solving

• Cloudflare-hosted
• Function calling
• Reasoning

[Meta logo]

detr-resnet-50

Beta

DEtection TRansformer (DETR) model trained end-to-end on COCO 2017 object detection (118k annotated images).

• Cloudflare-hosted

[HuggingFace logo]

distilbert-sst-2-int8

Distilled BERT model that was finetuned on SST-2 for sentiment classification

• Cloudflare-hosted

[Google logo]

embeddinggemma-300m

EmbeddingGemma is a 300M parameter, state-of-the-art for its size, open embedding model from Google, built from Gemma 3 (with T5Gemma initialization) and the same research and technology used to create Gemini models. EmbeddingGemma produces vector representations of text, making it well-suited for search and retrieval tasks, including classification, clustering, and semantic similarity search. This model was trained with data in 100+ spoken languages.

• Cloudflare-hosted

[Deepgram logo]

flux

Flux is the first conversational speech recognition model built specifically for voice agents.

• Cloudflare-hosted
• Partner
• Real-time

[Black Forest Labs logo]

flux-1-schnell

FLUX.1 [schnell] is a 12 billion parameter rectified flow transformer capable of generating images from text descriptions.

• Cloudflare-hosted

[Black Forest Labs logo]

flux-2-dev

FLUX.2 [dev] is an image model from Black Forest Labs where you can generate highly realistic and detailed images, with multi-reference support.

• Cloudflare-hosted
• Partner

[Black Forest Labs logo]

flux-2-klein-4b

FLUX.2 [klein] is an ultra-fast, distilled image model. It unifies image generation and editing in a single model, delivering state-of-the-art quality enabling interactive workflows, real-time previews, and latency-critical applications.

• Cloudflare-hosted
• Partner

[Black Forest Labs logo]

flux-2-klein-9b

FLUX.2 [klein] 9B is an ultra-fast, distilled image model with enhanced quality. It unifies image generation and editing in a single model, delivering state-of-the-art quality enabling interactive workflows, real-time previews, and latency-critical applications.

• Cloudflare-hosted
• Partner

[Google logo]

gemma-2b-it-lora

Beta

This is a Gemma-2B base model that Cloudflare dedicates for inference with LoRA adapters. Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.

• Cloudflare-hosted
• LoRA

[Google logo]

gemma-3-12b-it

Gemma 3 models are well-suited for a variety of text generation and image understanding tasks, including question answering, summarization, and reasoning. Gemma 3 models are multimodal, handling text and image input and generating text output, with a large, 128K context window, multilingual support in over 140 languages, and is available in more sizes than previous versions.

• Cloudflare-hosted
• LoRA
• Deprecated

[Google logo]

gemma-4-26b-a4b-it

Gemma 4 is Google's most intelligent family of open models, built from Gemini 3 research to maximize intelligence-per-parameter.

• Cloudflare-hosted
• Function calling
• Reasoning
• Vision

[Google logo]

gemma-7b-it

Beta

Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. They are text-to-text, decoder-only large language models, available in English, with open weights, pre-trained variants, and instruction-tuned variants.

• Cloudflare-hosted
• LoRA
• Deprecated

[Google logo]

gemma-7b-it-lora

Beta

This is a Gemma-7B base model that Cloudflare dedicates for inference with LoRA adapters. Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.

• Cloudflare-hosted
• LoRA

[Zhipu AI logo]

glm-4.7-flash

GLM-4.7-Flash is a fast and efficient multilingual text generation model with a 131,072 token context window. Optimized for dialogue, instruction-following, and multi-turn tool calling across 100+ languages.

• Cloudflare-hosted
• Function calling
• Reasoning

[Zhipu AI logo]

glm-5.2

Z.ai's flagship agentic coding model

• Cloudflare-hosted
• Function calling
• Reasoning

[OpenAI logo]

gpt-oss-120b

OpenAI's open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases gpt-oss-120b is for production, general purpose, high reasoning use-cases.

• Cloudflare-hosted
• Function calling
• Reasoning

[OpenAI logo]

gpt-oss-20b

OpenAI's open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases gpt-oss-20b is for lower latency, and local or specialized use-cases.

• Cloudflare-hosted
• Function calling
• Reasoning

[IBM logo]

granite-4.0-h-micro

Granite 4.0 instruct models deliver strong performance across benchmarks, achieving industry-leading results in key agentic tasks like instruction following and function calling. These efficiencies make the models well-suited for a wide range of use cases like retrieval-augmented generation (RAG), multi-agent workflows, and edge deployments.

• Cloudflare-hosted
• Function calling

[Moonshot AI logo]

kimi-k2.5

Kimi K2.5 is a frontier-scale open-source model with a 256k context window, multi-turn tool calling, vision inputs, and structured outputs for agentic workloads.

• Cloudflare-hosted
• Function calling
• Deprecated
• Reasoning
• Vision

[Moonshot AI logo]

kimi-k2.6

Kimi K2.6 is a frontier-scale open-source 1T parameter model with a 262.1k context window, multi-turn tool calling, vision inputs, and structured outputs for agentic workloads.

• Cloudflare-hosted
• Function calling
• Reasoning
• Vision

[Moonshot AI logo]

kimi-k2.7-code

Kimi K2.7 is a frontier-scale open-source 1T parameter model with a 262.1k context window, multi-turn tool calling, vision inputs, and structured outputs for agentic workloads.

• Cloudflare-hosted
• Function calling
• Reasoning
• Vision

[Meta logo]

llama-2-7b-chat-fp16

Full precision (fp16) generative text model with 7 billion parameters from Meta

• Cloudflare-hosted
• Deprecated

[Meta logo]

llama-2-7b-chat-hf-lora

Beta

This is a Llama2 base model that Cloudflare dedicated for inference with LoRA adapters. Llama 2 is a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. This is the repository for the 7B fine-tuned model, optimized for dialogue use cases and converted for the Hugging Face Transformers format.

• Cloudflare-hosted
• LoRA

[Meta logo]

llama-2-7b-chat-int8

Quantized (int8) generative text model with 7 billion parameters from Meta

• Cloudflare-hosted
• Deprecated

[Meta logo]

llama-3-8b-instruct

Generation over generation, Meta Llama 3 demonstrates state-of-the-art performance on a wide range of industry benchmarks and offers new capabilities, including improved reasoning.

• Cloudflare-hosted
• Deprecated

[Meta logo]

llama-3-8b-instruct-awq

Quantized (int4) generative text model with 8 billion parameters from Meta.

• Cloudflare-hosted
• Deprecated

[Meta logo]

llama-3.1-70b-instruct

The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models. The Llama 3.1 instruction tuned text only models are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.

• Cloudflare-hosted
• Deprecated

[Meta logo]

llama-3.1-8b-instruct

The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models. The Llama 3.1 instruction tuned text only models are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.

• Cloudflare-hosted
• Deprecated

[Meta logo]

llama-3.1-8b-instruct-awq

Quantized (int4) generative text model with 8 billion parameters from Meta.

• Cloudflare-hosted
• Deprecated

[Meta logo]

llama-3.1-8b-instruct-fast

[Fast version] The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models. The Llama 3.1 instruction tuned text only models are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.

• Cloudflare-hosted

[Meta logo]

llama-3.1-8b-instruct-fp8

Llama 3.1 8B quantized to FP8 precision

• Cloudflare-hosted

[Meta logo]

llama-3.2-11b-vision-instruct

The Llama 3.2-Vision instruction-tuned models are optimized for visual recognition, image reasoning, captioning, and answering general questions about an image.

• Cloudflare-hosted
• LoRA
• Vision

[Meta logo]

llama-3.2-1b-instruct

The Llama 3.2 instruction-tuned text only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarization tasks.

• Cloudflare-hosted

[Meta logo]

llama-3.2-3b-instruct

The Llama 3.2 instruction-tuned text only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarization tasks.

• Cloudflare-hosted

[Meta logo]

llama-3.3-70b-instruct-fp8-fast

Llama 3.3 70B quantized to fp8 precision, optimized to be faster.

• Cloudflare-hosted
• Batch
• Function calling

[Meta logo]

llama-4-scout-17b-16e-instruct

Meta's Llama 4 Scout is a 17 billion parameter model with 16 experts that is natively multimodal. These models leverage a mixture-of-experts architecture to offer industry-leading performance in text and image understanding.

• Cloudflare-hosted
• Batch
• Function calling
• Vision

[Meta logo]

llama-guard-3-8b

Llama Guard 3 is a Llama-3.1-8B pretrained model, fine-tuned for content safety classification. Similar to previous versions, it can be used to classify content in both LLM inputs (prompt classification) and in LLM responses (response classification). It acts as an LLM it generates text in its output that indicates whether a given prompt or response is safe or unsafe, and if unsafe, it also lists the content categories violated.

• Cloudflare-hosted
• LoRA

[Leonardo logo]

lucid-origin

Lucid Origin from Leonardo.AI is their most adaptable and prompt-responsive model to date. Whether you're generating images with sharp graphic design, stunning full-HD renders, or highly specific creative direction, it adheres closely to your prompts, renders text with accuracy, and supports a wide array of visual styles and aesthetics from stylized concept art to crisp product mockups.

• Cloudflare-hosted
• Partner

[Meta logo]

m2m100-1.2b

Multilingual encoder-decoder (seq-to-seq) model trained for Many-to-Many multilingual translation

• Cloudflare-hosted
• Batch

[MyShell logo]

melotts

MeloTTS is a high-quality multi-lingual text-to-speech library by MyShell.ai.

• Cloudflare-hosted

[Meta logo]

meta-llama-3-8b-instruct

Generation over generation, Meta Llama 3 demonstrates state-of-the-art performance on a wide range of industry benchmarks and offers new capabilities, including improved reasoning.

• Cloudflare-hosted
• Deprecated

[MistralAI logo]

mistral-7b-instruct-v0.1

Instruct fine-tuned version of the Mistral-7b generative text model with 7 billion parameters

• Cloudflare-hosted
• LoRA
• Deprecated

[MistralAI logo]

mistral-7b-instruct-v0.2

Beta

The Mistral-7B-Instruct-v0.2 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-7B-v0.2. Mistral-7B-v0.2 has the following changes compared to Mistral-7B-v0.1: 32k context window (vs 8k context in v0.1), rope-theta = 1e6, and no Sliding-Window Attention.

• Cloudflare-hosted
• LoRA
• Deprecated

[MistralAI logo]

mistral-7b-instruct-v0.2-lora

Beta

The Mistral-7B-Instruct-v0.2 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-7B-v0.2.

• Cloudflare-hosted
• LoRA

[MistralAI logo]

mistral-small-3.1-24b-instruct

Building upon Mistral Small 3 (2501), Mistral Small 3.1 (2503) adds state-of-the-art vision understanding and enhances long context capabilities up to 128k tokens without compromising text performance. With 24 billion parameters, this model achieves top-tier capabilities in both text and vision tasks.

• Cloudflare-hosted
• Function calling

[NVIDIA logo]

nemotron-3-120b-a12b

NVIDIA Nemotron 3 Super is a hybrid MoE model with leading accuracy for multi-agent applications and specialized agentic AI systems.

• Cloudflare-hosted
• Function calling
• Reasoning

[Deepgram logo]

nova-3

Transcribe audio using Deepgrams speech-to-text model

• Cloudflare-hosted
• Batch
• Partner
• Real-time

[Microsoft logo]

phi-2

Beta

Phi-2 is a Transformer-based model with a next-word prediction objective, trained on 1.4T tokens from multiple passes on a mixture of Synthetic and Web datasets for NLP and coding.

• Cloudflare-hosted
• Deprecated

[Leonardo logo]

phoenix-1.0

Phoenix 1.0 is a model by Leonardo.Ai that generates images with exceptional prompt adherence and coherent text.

• Cloudflare-hosted
• Partner

[Qwen logo]

qwen2.5-coder-32b-instruct

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5:

• Cloudflare-hosted
• LoRA

[Qwen logo]

qwen3-30b-a3b-fp8

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support.

• Cloudflare-hosted
• Batch
• Function calling
• Reasoning

[Qwen logo]

qwen3-embedding-0.6b

The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.

• Cloudflare-hosted

[Qwen logo]

qwen3.8-27b

Qwen 3.8 27B is a 27-billion-parameter instruction-tuned language model from Alibaba's Qwen family, designed for vision, efficient general-purpose text generation and agentic workloads.

• Cloudflare-hosted
• Function calling
• Reasoning
• Vision

[Qwen logo]

qwq-32b

QwQ is the reasoning model of the Qwen series. Compared with conventional instruction-tuned models, QwQ, which is capable of thinking and reasoning, can achieve significantly enhanced performance in downstream tasks, especially hard problems. QwQ-32B is the medium-sized reasoning model, which is capable of achieving competitive performance against state-of-the-art reasoning models, e.g., DeepSeek-R1, o1-mini.

• Cloudflare-hosted
• LoRA
• Reasoning

[Microsoft logo]

resnet-50

50 layers deep image classification CNN trained on more than 1M images from ImageNet

• Cloudflare-hosted

[Pipecat logo]

smart-turn-v2

An open source, community-driven, native audio turn detection model in 2nd version

• Cloudflare-hosted
• Batch
• Real-time

[Defog logo]

sqlcoder-7b-2

Beta

This model is intended to be used by non-technical users to understand data inside their SQL databases.

• Cloudflare-hosted
• Deprecated

[RunwayML logo]

stable-diffusion-v1-5-img2img

Beta

Stable Diffusion is a latent text-to-image diffusion model capable of generating photo-realistic images. Img2img generate a new image from an input image with Stable Diffusion.

• Cloudflare-hosted

[RunwayML logo]

stable-diffusion-v1-5-inpainting

Beta

Stable Diffusion Inpainting is a latent text-to-image diffusion model capable of generating photo-realistic images given any text input, with the extra capability of inpainting the pictures by using a mask.

• Cloudflare-hosted

[Stability.ai logo]

stable-diffusion-xl-base-1.0

Beta

Diffusion-based text-to-image generative model by Stability AI. Generates and modify images based on text prompts.

• Cloudflare-hosted

[ByteDance logo]

stable-diffusion-xl-lightning

Beta

SDXL-Lightning is a lightning-fast text-to-image generation model. It can generate high-quality 1024px images in a few steps.

• Cloudflare-hosted

[Unum logo]

uform-gen2-qwen-500m

Beta

UForm-Gen is a small generative vision-language model primarily designed for Image Captioning and Visual Question Answering. The model was pre-trained on the internal image captioning dataset and fine-tuned on public instructions datasets: SVIT, LVIS, VQAs datasets.

• Cloudflare-hosted
• Deprecated

[OpenAI logo]

whisper

Whisper is a general-purpose speech recognition model. It is trained on a large dataset of diverse audio and is also a multitasking model that can perform multilingual speech recognition, speech translation, and language identification.

• Cloudflare-hosted

[OpenAI logo]

whisper-large-v3-turbo

Whisper is a pre-trained model for automatic speech recognition (ASR) and speech translation.

• Cloudflare-hosted
• Batch

[OpenAI logo]

whisper-tiny-en

Beta

Whisper is a pre-trained model for automatic speech recognition (ASR) and speech translation. Trained on 680k hours of labelled data, Whisper models demonstrate a strong ability to generalize to many datasets and domains without the need for fine-tuning. This is the English-only version of the Whisper Tiny model which was trained on the task of speech recognition.

• Cloudflare-hosted