Models

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.

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).

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.

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.

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).

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.

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.

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.

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.

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.

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.

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.

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.

wan-3.0

Alibaba's Wan 3.0 text-to-video model. Generates cinematic videos from text prompts with adaptive aspect ratio, 480P, 720P, or 1080P resolution, and configurable duration.

wan-3.0-prime

Alibaba's Wan 3.0 Prime text-to-video model. Generates cinematic videos from text prompts with adaptive aspect ratio, 480P, 720P, or 1080P resolution, and configurable duration.

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.

claude-fable-5.1

Claude Fable 5.1 is Anthropic's next model in the Fable family, with improvements in agentic coding, long-running agentic workflows, knowledge work, front-end and visual code generation, and finance and analysis tasks. It supports adaptive thinking and a 1M token context window.

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.

claude-haiku-5.5

Claude Haiku 5.5 is a fast, cost-efficient model for high-volume, latency-sensitive tasks such as classification, extraction, routing, and subagent work.

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.

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.

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.

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.

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.

claude-opus-5.5

Claude Opus 5.5 is Anthropic's model for long-running agentic coding and knowledge work. It combines a one million token context window with adaptive thinking that is always on, a 128K token maximum output, and a medium default effort level.

claude-sonnet-4.5

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

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.

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.

claude-sonnet-5.5

Claude Sonnet 5.5 offers the best combination of speed and intelligence, with adaptive thinking for coding, tool use, reasoning, and long-horizon work.

universal-3-pro

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

universal-3.5-pro

AssemblyAI's Universal-3.5 Pro speech recognition model for fast, high-accuracy transcription.

flux-1-kontext-max

FLUX.1 Kontext [max] is Black Forest Labs' highest-quality Kontext model for text-to-image generation and context-aware image editing.

flux-1-kontext-pro

FLUX.1 Kontext [pro] creates and edits images from text prompts with strong character and style consistency.

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.

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.

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.

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.

flux-video-upscale

FLUX Video Upscale increases video resolution with a precise mode for source-faithful results and a creative mode for stronger detail enhancement. It accepts clips up to 20 seconds and preserves audio.

FIBO Edit changes one to four input images from a natural-language or structured JSON (VGL) instruction. Add a black-and-white mask to edit only part of a single image, or refer to several images as "image 1", "image 2", and so on to combine them or transfer a style.

FIBO Generate creates images from a text prompt, a reference image, or both. Each result includes the seed and the structured JSON (VGL) prompt used to render it; send them back to recreate the image exactly or to refine it with a new prompt. Outputs are 1MP or 4MP in nine aspect ratios.

Bria's RMBG 2.0 model removes the background from an image and returns a PNG cutout with a transparent background. Partial transparency from the input's alpha channel is kept by default.

Bria's video background removal replaces the background of a clip up to 60 seconds long with a solid color or with transparency. Transparent output needs the mov_proresks (ProRes) preset. Frame rate and audio are preserved, and the output resolution matches the input unless auto_zoom crops to the subject.

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.

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.

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.

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.

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.

seedream-4.5

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

seedream-5-lite

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

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.

DeepSeek V4 Pro is a high-capability reasoning model from DeepSeek, served via Fireworks infrastructure for production-grade inference.

eleven-flash-v2-5

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

eleven-multilingual-v2

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

eleven-turbo-v2-5

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

eleven-v3

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

music-v2

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

gemini-2.5-flash

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

gemini-2.5-flash-lite

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

gemini-2.5-pro

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

gemini-3-flash

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

gemini-3.1-flash-lite

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

gemini-3.1-pro

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

gemini-3.5-flash

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

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.

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.

gemini-3.7-flash

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

gemini-3.8-flash

Our most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows.

gemini-3.8-flash-lite-tts

Gemini 3.8 Flash-Lite TTS is a fast, cost-efficient text-to-speech model optimized for high-throughput production workloads.

gemini-3.8-flash-tts

Gemini 3.8 Flash TTS is Google's flagship text-to-speech model for expressive, natural speech generation.

gemini-omni-1.1-flash

High-performance multimodal video generation and editing model with conversational controls and generated audio.

gemini-omni-flash

Preview high-performance multimodal video generation and editing model with conversational controls and generated audio.

nano-banana

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

nano-banana-2

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

nano-banana-2-lite

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

nano-banana-2.1

Google's Nano Banana 2.1 image generation and editing model, optimized for a balance of quality and performance.

nano-banana-pro

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

veo-3.1

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

veo-3.1-fast

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

tts-1.5-max

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

tts-1.5-mini

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

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.

More than 2x the size of Medium, with softer post-training. Outputs are rawer, more textured, and more flexible — at its best, Large produces results Medium can't match. Strongest on photorealism, raw looks (motion blur, grain, low dynamic range), and expressive and artistic styles.

Smaller, faster, more cost-efficient. Extensive post-training makes outputs especially stable and consistent across generations. Strongest on illustration, anime, painting, and other expressive or artistic styles.

The fastest Krea 2 model, built for low-cost iteration on expressive illustrations, style-driven concepts, and rapid visual exploration. Keeps the Krea 2 style system and expressive visual range but uses a distilled sampling schedule so you can move through ideas much faster. Especially useful for expressive illustration, graphic styles, typography experiments, and quick campaign or concept directions.

Lightricks LTX-2.5 Fast is a fast video generation model for text-to-video and image-to-video workflows, with synchronized audio, configurable duration, resolution, and frame rate.

muse-spark-1.1

Muse Spark 1.1 is Meta's agentic and coding model for multi-step tool loops, software engineering assistants, and long-context reasoning.

muse-spark-1.2

Muse Spark 1.2 is Meta's agentic and coding model for multi-step tool loops, software engineering assistants, and long-context reasoning.

muse-spark-1.3

Muse Spark 1.3 is Meta's latest agentic and coding model, tuned for multi-step tool loops, browser tasks, software engineering, and long-horizon reasoning.

muse-wonder-1.0

Muse Wonder 1.0 is Meta's agentic and coding model for complex software engineering, multi-step tool loops, and long-context reasoning.

h3

A multimodal video generation model supporting text-to-video, first and last frame image-to-video, and reference-to-video generation with 768P and 2K output.

h3-max

A fast multimodal video generation model supporting text-to-video and image-to-video generation at 480P and 768P.

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.

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.

m2.7

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

m3

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

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.

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.

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.

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.

gpt-4.1

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

gpt-4.1-mini

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

gpt-4.1-nano

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

gpt-4o

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

gpt-4o-mini

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

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.

gpt-5

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

gpt-5-mini

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

gpt-5-nano

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

gpt-5.1

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

gpt-5.4

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

gpt-5.4-mini

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

gpt-5.4-nano

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

gpt-5.4-pro

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

gpt-5.5

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

gpt-5.5-pro

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

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.

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.

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.

gpt-6-astra

GPT-6 Astra is OpenAI's most capable model, built for complex reasoning, coding, computer use, research, and document creation.

gpt-6-luna

GPT-6 Luna is OpenAI's most efficient GPT-6 model, built for focused, high-volume tasks.

gpt-6-sol

GPT-6 Sol is OpenAI's mid-tier GPT-6 model, built to power complex coding and agentic workflows.

gpt-6.1-sol

GPT-6.1 Sol is OpenAI's newest Sol model, delivering near-Astra performance at a lower cost for complex coding, computer use, and professional work.

gpt-image-1.5

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

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.

gpt-image-2.5-flare

OpenAI's fastest high-quality everyday image generation model. It accepts text and image inputs and produces images with low, medium, high, xhigh, max, and auto quality settings.

gpt-image-2.5-sunburst

OpenAI's most capable image generation and editing model. It accepts text and image inputs and produces images with low, medium, high, xhigh, max, and auto quality settings.

o3

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

o3-mini

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

o4-mini

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

tts-1

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

tts-1-hd

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

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

recraftv4-1-vector

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

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.

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.

recraftv4-vector

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

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.

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.

Inkling is Thinking Machines' open-weights hybrid reasoning model, built on a mixture-of-experts architecture. It reasons by default, exposing its chain-of-thought as leading thinking content blocks, with reasoning effort tunable via a Tinker-specific output_config.effort parameter. Available through Tinker's beta Anthropic Messages-compatible endpoint alongside tool use, streaming, and multi-turn conversations. Currently intended for low-traffic testing and internal use rather than high-throughput production deployments; prompt caching, citations, and audio input are not supported through this endpoint.

The 256K-context variant of Inkling, Thinking Machines' open-weights hybrid reasoning MoE model. Same hybrid reasoning, tool-use, and streaming support as the base model, with an extended context window for longer conversations and documents. Currently intended for low-traffic testing and internal use rather than high-throughput production deployments.

Topaz Labs Starlight Fast 2 (slf-2) enhances and upscales source video to 1080p or 4K with faster processing than Starlight Precise. It is designed for video restoration workflows where turnaround time is important.

Topaz Labs Starlight Precise 2.6 (slp-2.6) is a video enhancement model that upscales and restores low-quality source video to 1080p or 4K. The quality-focused variant of the Starlight family, prioritizing detail reconstruction over speed. Billed using provider-reported processing credits.

Jev is TypeSafe's structured evaluation model. It evaluates one state against typed Noul, Choice, and Score questions and returns calibrated answers with probabilities and confidence.

Pareto is Unbiased's blended AI model. It engages multiple language models in parallel for each request, synthesizes one answer, and supports text and vision inputs through a single API response.

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.

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.

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.

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.

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.

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).

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).

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.

grok-4.7

xAI's Grok 4.7, a frontier reasoning model 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, xhigh).

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.

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.

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.

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).

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.

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.

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.

grok-voice

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

Apertus 1.5 is an 8B parameter language model designed to advance the state of multilingual, multimodal, fully open, and transparent AI. The models support a wide range of languages, handle contexts of up to 262,144 tokens, and it uses only fully open training data whilst delivering performance comparable to other models of similar size. For access, please fill out this form: https://forms.gle/kgvmkr6ucHN3xNuH6

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.

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.

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.

bge-base-en-v1.5

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

bge-large-en-v1.5

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

bge-m3

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

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.

bge-small-en-v1.5

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

clef

Clef is a 27B multimodal decision model that turns a state and a schema of typed questions into decisions. It reads the state as text, JSON, images, or video, and returns a probability for every allowed option of every question.

clef-flash

Clef-flash is a fast 9B multimodal decision model that turns a state and a schema of typed questions into decisions. It reads the state as text, JSON, images, or video, and returns a probability for every allowed option of every question.

clef-omni

Clef-omni is a multimodal decision model built on a 30B-parameter mixture-of-experts (3B active) backbone. It turns a state and a schema of typed questions into decisions, reads the state as text, JSON, images, audio, or video, and returns a probability for every allowed option of every question.

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.

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.

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

distilbert-sst-2-int8

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

Stable Diffusion model that has been fine-tuned to be better at photorealism without sacrificing range.

embeddinggemma-300m

Beta

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.

EuroLLM-9B is a 9B parameter model trained on 4 trillion tokens divided across the considered languages and several data sources: Web data, parallel data (en-xx and xx-en), and high-quality datasets. For access, please fill out this form: https://forms.gle/kgvmkr6ucHN3xNuH6

flux

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

flux-1-schnell

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

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.

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.

flux-2-klein-9b

FLUX.2 [klein] 9B is a 9 billion parameter model that can generate images from text descriptions and supports multi-reference editing capabilities.

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.

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.

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.

SEA-LION stands for Southeast Asian Languages In One Network, which is a collection of Large Language Models (LLMs) which have been pretrained and instruct-tuned for the Southeast Asia (SEA) region.

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.

glm-5.2

Z.ai's flagship agentic coding model

glm-5.3

GLM-5.3 is Z.ai's flagship agentic coding model, pairing a 1M-token context window with reasoning, function calling, and structured outputs to power multi-step, tool-driven development workflows.

glm-5.3-flash

The first natively multimodal model in the GLM-5 series. With 320B total parameters and just 18B active parameters, it outperforms GLM-5.2 across benchmarks and real-world workloads at one-tenth the price, while approaching Claude Opus 4.8 on coding and agentic benchmarks.

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.

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.

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.

IndicTrans2 is the first open-source transformer-based multilingual NMT model that supports high-quality translations across all the 22 scheduled Indic languages

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.

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.

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.

llama-3.1-8b-instruct-fp8

Llama 3.1 8B quantized to FP8 precision

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.

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.

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.

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

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

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.

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.

LLaVA is an open-source chatbot trained by fine-tuning LLaMA/Vicuna on GPT-generated multimodal instruction-following data. It is an auto-regressive language model, based on the transformer architecture.

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.

m2m100-1.2b

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

melotts

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

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

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

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.

Moondream 3 is a fast, efficient 9B mixture-of-experts vision language model (2B active parameters) that delivers frontier-level visual reasoning for tasks like object detection, pointing, OCR, and structured output.

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.

nova-3

Transcribe audio using Deepgram’s speech-to-text model

phoenix-1.0

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

PLaMo-Embedding-1B is a Japanese text embedding model developed by Preferred Networks, Inc. It can convert Japanese text input into numerical vectors and can be used for a wide range of applications, including information retrieval, text classification, and clustering.

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:

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.

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.

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.

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.

resnet-50

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

smart-turn-v2

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

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.

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.

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.

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.

whisper-large-v3-turbo

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

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.