AI Models

Multilingual TTS with voice cloning from 6-second samples. 17 languages. Very lightweight.

8x faster than Large with minimal quality loss. Best speed/accuracy trade-off.

The original Stable Diffusion. Huge ecosystem of LoRAs, ControlNets, and extensions. Best compatibility.

Versatile text-to-speech. Generates speech, music, and sound effects. Supports voice cloning.

Ultra-compact 1B model. Runs on virtually any device including smartphones.

Lightweight video generation. Consumer GPU friendly with CPU offloading.

AI music generation from text prompts. Creates 30-second clips of genre-specific music.

Best accuracy speech-to-text. Supports 99 languages. The gold standard for transcription.

Accessible video generation model. Creates short video clips from text descriptions.

Meta's compact 3B model designed for edge and mobile deployment.

Higher resolution (1024x1024) image generation. Better detail and prompt following than SD 1.5.

Tiny but capable 3.8B model. Runs on almost any hardware including phones.

Higher quality video generation. Better motion and detail than the 2B variant.

Lightweight vision-language model. Analyze images on consumer hardware.

Multimodal model supporting image, video, and text input. Dynamic resolution handling.

Efficient 7B model from Mistral AI with strong performance for its size.

Efficient 7B model with strong coding and reasoning abilities. Great for local deployment.

Efficient coding model with 131K context. Great local Copilot replacement.

Meta's 8B parameter instruction-tuned model. Great balance of performance and efficiency for local deployment.

Compact reasoning model. Good reasoning capabilities in a small package.

Latest SD with MMDiT architecture. Best prompt adherence with T5-XXL encoder.

Google's efficient 9B model. Great performance-to-size ratio.

High-quality text-to-video. Produces realistic motion but requires significant VRAM.

State-of-the-art image generation. 12B parameters. Best open-source image quality available.

Fast version of Flux. Only 4 inference steps needed. Near-instant generation on good GPUs.

Microsoft's 14B parameter model. Punches well above its weight class on reasoning benchmarks.

Lightweight MoE code model. Strong coding with only 2.4B active parameters.

Mistral's dedicated code generation model. Supports 80+ programming languages.

Google's 27B instruction-tuned model. Strong general performance with efficient architecture.

Specialized coding model. One of the best open-source coding assistants available.

Reasoning-focused model distilled from DeepSeek R1. Excellent at complex problem-solving.

Vision-language model. Analyze images and answer questions about them. Best quality variant.

Mixture-of-experts model with 8 experts of 7B each. Excellent quality at moderate resource requirements.

Meta's flagship 70B parameter model. Excellent performance rivaling GPT-4 on many benchmarks.

Latest Meta 70B model with improved reasoning and multilingual capabilities.

Alibaba's top-tier 72B model. Excellent at coding, math, and multilingual tasks.

Cohere's 104B parameter model. Excellent for RAG, tool use, and enterprise applications.

DeepSeek's latest flagship model with 685B total parameters (37B active via MoE). State-of-the-art performance.

Fastest Whisper model. Runs on anything including phones. 10x faster than Large.

Fast transcription for resource-limited devices. 4x faster than Large.

Lightweight music generation. Quick and accessible on most hardware.

Animate existing Stable Diffusion images. Lightweight motion module on top of SD 1.5.

Good balance of speed and accuracy. 2x faster than Large, slight quality trade-off.

Trained on The Stack v2 with 600+ languages. Strong code completion.

Strong 34B model from 01.AI with good multilingual and reasoning performance.