Overview / start here. For the full catalog of model families see Supported Models; for connecting and configuring each provider see Providers; for the desktop download panel see Models Manager.
Run models locally, through cloud APIs with your own keys, or both in the same graph.
Local vs. cloud
Local
- Data stays on your disk
- Free after download
- Works offline
- 4–20 GB per model; needs a capable GPU or Apple Silicon
Cloud (BYOK)
- No download
- Runs on the provider’s hardware
- Latest model releases
- Billed by the provider, at the provider’s price — no NodeTool markup
- Requires internet; data goes to the provider
Mixed
Pick the best provider per node:
- ASR (Whisper) — local for sensitive audio
- Image generation — Flux locally for control, FAL/KIE cloud for speed
- Document processing — local for confidential files
Cloud models
NodeTool reaches thousands of cloud models across 30+ providers, all through the same generic nodes. The catalog below is grouped by what you’re generating — text, images, video, speech and music, 3D, embeddings. Every model runs on your own API key (BYOK): you’re billed by the provider at the provider’s price, with no NodeTool markup. Add a key in Settings → Models & Providers and the models show up in the node’s model dropdown.
Most chat providers fetch their model list live from the provider’s API, so the list reflects new releases. A few use fixed lists, such as kie.ai’s chat gateway and Meta AI’s offline fallback. Image, video, audio, and 3D models come from manifests or static lists that each provider ships (packages/runtime/src/providers/ and the packages/*-nodes/ packages), so they track what NodeTool exposes. The model names below are examples and drift as providers release.
Text & chat models (LLMs)
The generic nodetool.agents.Agent and chat nodes route to whichever provider owns the model you pick. Swap claude-opus-5 for gpt-5.6 or gemini-3.1-pro-preview without changing the rest of the graph.
| Provider | Model families |
|---|---|
 OpenAI |
GPT-5.6 (plus the Sol, Terra, and Luna variants), GPT-5.5, GPT-5.5 Pro, GPT-5.4 with its Pro/mini/nano tiers, GPT-5, GPT-5 mini, GPT-5 nano |
 Anthropic |
Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Opus 4.8/4.7/4.6, Claude Sonnet 4.6, Claude Haiku 4.5 |
 Google Gemini |
Gemini 3.5 Flash, Gemini 3.1 Pro, Gemini 3.1 Flash-Lite, Gemini 2.5 Pro |
 xAI |
Grok 4, Grok 3, Grok Code |
| DeepSeek | DeepSeek-V3, DeepSeek-R1 (reasoning) |
| Groq | Llama, Qwen, GPT-OSS, Kimi K2 on LPU inference |
| Mistral | Mistral Large, Mistral Medium, Mistral Small, Mixtral, Codestral, Magistral |
| Cerebras | Llama, Qwen, GPT-OSS on wafer-scale inference |
| Alibaba Cloud | Qwen3-Max, Qwen-Plus, Qwen-Flash, Qwen-Turbo, Qwen3-VL |
| GMI Cloud | Llama, DeepSeek, Qwen (open-weight) |
| Moonshot | Kimi K2, Kimi latest |
| Meta AI | Muse Spark |
 MiniMax |
MiniMax M2.7, M2.5, M2.1, M2 (plus high-speed variants) |
| OpenRouter | 300+ models proxied through one key (Claude, GPT, Gemini, Llama, Qwen, DeepSeek, …) |
| Requesty | Claude, GPT, Gemini, DeepSeek, Grok and more through one OpenAI-compatible router |
| API Route | Hosted Claude, GPT, Gemini and DeepSeek chat models through an OpenAI-compatible API |
| Opper | Claude, GPT, Gemini, DeepSeek, Kimi and more through one EU-hosted OpenAI-compatible gateway |
| Together AI | Llama, Qwen, DeepSeek, Mixtral, GLM, Kimi, and more open models |
| Evolink | GPT, Claude, Gemini, DeepSeek through one gateway key |
| kie.ai | GPT, Claude, Grok, Kimi, DeepSeek, and Gemini models through its chat gateway |
| Codex (OpenAI OAuth) | GPT chat models via your ChatGPT/Codex login |
| Claude Agent SDK | Claude via your local claude CLI subscription |
| Ollama · vLLM · LM Studio · llama.cpp · llama.cpp local | Any local open-weight model served by your own engine (Llama, Qwen, Gemma, DeepSeek, Mistral, Phi, GPT-OSS) |
| HuggingFace | The most-liked warm inference models on the Hub, through Inference Providers |
| FAL | Chat through fal’s OpenAI-compatible route (OpenRouter’s catalog) |
Reasoning, tool calling, and vision input are exposed where the provider supports them. For local text generation without an API key, see Supported Models (Ollama, llama.cpp, MLX, Transformers.js).
Image generation models
Generate and edit images through nodetool.image.TextToImage and nodetool.image.ImageToImage.
Flagship models
| Model | Provider | Capabilities | Key features |
|---|---|---|---|
| Black Forest Labs FLUX.2 |  BFL |
T2I with control | Photoreal images, multi-reference consistency, accurate text, flexible control |
| Google Nano Banana 2.0 / Pro | Google |
High-res T2I/Edit | 2K output, 4K upscaling, better text and character consistency |
| GPT Image 2 | OpenAI |
T2I/Edit | Photorealistic generation and instruction-based editing |
| Ideogram V3 | Ideogram | T2I/Edit | Typography rendering and artistic style control |
| Seedream 4.5 | ByteDance | T2I/Edit | High-fidelity generation and instruction-based editing |
| Z-Image Turbo | Z-AI | T2I | Fast generation with strong prompt adherence |
Full catalog by provider
| Provider | Image models |
|---|---|
| OpenAI |
GPT Image 2, GPT Image 1.5, GPT Image 1, GPT Image 1 mini |
| Google |
Gemini 3.1 Flash Image (Nano Banana 2), Gemini 3 Pro Image (Nano Banana Pro) |
| Black Forest Labs |
FLUX.1 [schnell], FLUX.1 [dev], FLUX.1 Pro, FLUX 1.1 Pro, FLUX 1.1 Pro Ultra, FLUX.2 [dev/flex/pro/max], FLUX.2 Klein, FLUX Kontext [pro/max], FLUX Fill, FLUX Canny, FLUX Depth, FLUX Redux, FLUX PuLID |
| Stability AI | Stable Diffusion 1.5, SD 2.1, SDXL, SDXL Lightning, Stable Diffusion 3 Medium, SD 3.5 [medium/large/large turbo], Stable Cascade |
| Qwen | Qwen Image, Qwen Image Edit (2509/2511/Plus), Qwen Image Max, Qwen Image Layered |
| ByteDance | Seedream 3.0, Seedream 4.0, Seedream 4.5, Seedream 5 Lite |
| Ideogram | Ideogram V2, V2 Turbo, V2a, V3 [balanced/quality/turbo], Ideogram Character |
| Recraft | Recraft V3, Recraft 20B, Recraft V4, Recraft V4 Pro (raster + SVG) |
| Others | Kolors, HiDream, Hunyuan Image 3, OmniGen v1/v2, Sana, PixArt-Σ, Aura Flow, CogView4, Luma Photon, GLM Image, Reve, Bria, ERNIE Image, Emu 3.5, Lumina, F-Lite, Fibo, Playground v2.5, Juggernaut, DreamShaper, Proteus, Recraft, Chroma, Janus, MiniMax Image-01, xAI Grok Imagine |
| Upscale & restore | Topaz, Real-ESRGAN, ESRGAN, SUPIR, Clarity Upscaler, SeedVR, GFPGAN, CodeFormer, AuraSR, SwinIR, DDColor |
Access aggregators (FAL, Replicate, kie.ai, and AtlasCloud) carry most of these families, plus community fine-tunes, LoRAs, and control variants.
Video generation models
Generate video through nodetool.video.TextToVideo and nodetool.video.ImageToVideo.
Flagship models
| Model | Provider | Capabilities | Key features |
|---|---|---|---|
| OpenAI Sora 2 Pro | OpenAI |
T2V/I2V up to 15s | Realistic motion, refined physics, synchronized native audio, 1080p |
| Google Veo 3.1 | Google |
T2V/I2V with references | Multi-image references, extended length, native 1080p with synced audio |
| ByteDance Seedance 2.0 | ByteDance | T2V/I2V | Cinematic video with stable characters and smooth motion |
| Runway Gen-4 / Aleph | Runway | T2V/I2V/Extend | Precise motion control and high fidelity |
| Luma Ray 2 | Luma AI | T2V/I2V/edit | Fast generation and creative editing |
| xAI Grok Imagine | xAI |
T2V/I2V/T2I | Coherent motion with synchronized audio |
| Alibaba Wan 2.6 |  Alibaba |
Multi-shot T2V/I2V | Affordable 1080p, stable characters, native audio |
| MiniMax Hailuo 2.3 | MiniMax |
High-fidelity T2V/I2V | Expressive characters, complex motion and lighting |
| Kling 3.0 |  Kling |
T2V/I2V with audio | Synchronized speech, ambient sound, and effects |
Full catalog by provider
| Provider | Video models |
|---|---|
| OpenAI |
Sora 2, Sora 2 Pro |
| Google |
Veo 3.1, Veo 3.1 Fast, Veo 3.1 Lite, Gemini Omni Flash |
| Kling |
Kling 2.1 [standard/pro/master], Kling 2.5 Turbo Pro, Kling 2.6, Kling 3.0, Kling Avatar, Kling Lip Sync |
| ByteDance | Seedance 1.0 [lite/pro], Seedance 1.5 Pro, Seedance 2.0 |
| Alibaba |
Wan 2.1, Wan 2.2, Wan 2.5, Wan 2.6, Wan VACE |
| MiniMax |
Hailuo 02, Hailuo 2.3, Video-01 [director/live] |
| Runway | Gen-3 Alpha, Gen-4, Gen-4 Turbo, Gen-4.5, Aleph |
| Luma AI | Dream Machine, Ray 2 [540p/720p], Ray Flash 2 |
| PixVerse | PixVerse v4, v4.5, v5, v5.6, v6 |
| Vidu | Vidu 2.0, Vidu Q1 |
| Others | LTX Video, LTX-2, Hunyuan Video (+ v1.5, Avatar, Foley), Mochi v1, CogVideoX-5B, Marey, Lucy, Magi, Kandinsky 5, Sana Video, SkyReels, Stable Video Diffusion, Zeroscope, Pika, xAI Grok Imagine |
| Lip sync & avatars | LatentSync, Lipsync 2 [pro], MultiTalk, OmniHuman, SadTalker, Live Portrait, Stable Avatar, DreamActor, InfiniteTalk |
| Upscale | Topaz Video, SeedVR, Crystal Video Upscaler |
Speech, audio & music models
Generate speech through nodetool.audio.TextToSpeech, transcribe with nodetool.text.AutomaticSpeechRecognition, and create music through provider music nodes.
Text-to-speech (TTS)
| Provider | TTS models |
|---|---|
| OpenAI |
TTS 1, TTS 1 HD, gpt-4o-mini-tts |
 ElevenLabs |
Eleven v3, Turbo v2.5, Flash v2.5, Multilingual v2 |
| Google |
Gemini TTS, Gemini 3.1 Flash TTS |
| MiniMax |
Speech 02 [hd/turbo], Speech 2.6, Speech 2.8 |
| Open models | Kokoro, Orpheus, Chatterbox [HD/Pro/Turbo/Multilingual], Dia, F5-TTS, XTTS v2, VibeVoice, Zonos, CSM-1B, Index-TTS 2, Qwen 3 TTS, Maya, StyleTTS2, Parler-TTS, Tortoise, VoiceCraft, OpenVoice, Cartesia Sonic |
Speech recognition (ASR)
| Provider | ASR models |
|---|---|
| OpenAI |
Whisper |
| Google |
Gemini audio transcription |
| ElevenLabs |
Scribe (speech-to-text) |
| Together AI | Whisper Large v3, Voxtral, Parakeet |
| HuggingFace | Whisper and other ASR models, plus MLX Whisper locally |
Music & sound
| Provider | Music & sound models |
|---|---|
| Suno (via kie.ai) | Suno — song generation, extend, cover, remix |
| ElevenLabs |
V3 Dialogue, Sound Effects |
| MiniMax |
Music 01, Music 1.5, Music 2.6 |
| Google |
Lyria 3.5, Lyria 3 Clip |
| Open models | MusicGen, Stable Audio 2.5, ACE-Step, DiffRhythm, YuE, Riffusion, Flux Music, MMAudio, ThinkSound |
3D generation models
Generate 3D assets through nodetool.model3d.TextTo3D and nodetool.model3d.ImageTo3D.
| Model | Provider | Capabilities | Key features |
|---|---|---|---|
| Meshy AI | Meshy (MESHY_API_KEY) |
T2M/I2M | Textured mesh generation |
| Rodin AI | Rodin (RODIN_API_KEY) |
T2M/I2M | High-fidelity 3D creation |
AtlasCloud also lists 3D models. Open 3D families such as Hunyuan3D and Trellis run through provider-specific FAL and Replicate nodes, not the generic nodes. See Providers for details.
Embedding models
Power RAG and semantic search through embedding nodes.
| Provider | Embedding models |
|---|---|
| OpenAI |
text-embedding-3-small, text-embedding-3-large, ada-002 |
| Google |
gemini-embedding-2, gemini-embedding-001 |
| Mistral | mistral-embed |
| Cohere | embed-v4.0, embed-english-v3.0, embed-multilingual-v3.0 |
| Voyage AI | voyage-3.5 and the Voyage line |
| Jina AI | jina-embeddings-v3 |
| Together AI | m2-bert, BGE, and other open embedding models |
| Ollama · llama.cpp local · Transformers.js | Embedding models that run on your machine with no key (Transformers.js downloads ONNX models on first use) |
| vLLM · OpenRouter · Replicate · HuggingFace | Any embedding model the server or provider lists |
Using these models
Access these models through NodeTool’s generic nodes:
- For Video: Use
nodetool.video.TextToVideoornodetool.video.ImageToVideo - For Images: Use
nodetool.image.TextToImage - For 3D: Use
nodetool.model3d.TextTo3Dornodetool.model3d.ImageTo3D - For Music: Use the kie.ai Suno nodes (Generate Music, Extend Music, Upload And Cover Audio, Add Vocals, and others)
- Select Provider: Click the model dropdown in the node properties
- Configure API: Add provider API keys in
Settings → Models & Providers
Access via kie.ai (recommended for broad model support): Many of these models are available through kie.ai, an AI provider aggregator that often offers competitive or lower pricing compared to upstream providers.
- Configure using
KIE_API_KEYinSettings → Models & Providers
Access via fal.ai:
- Configure using
FAL_API_KEYinSettings → Models & Providers
Cost Considerations: Cloud models typically charge per generation. Check each provider’s pricing before extensive use. Local models are free after download but require capable hardware.
Getting Started
Option 1: Start with Local Models (Recommended)
- Open the logo menu and choose Model Manager
- On an empty install it opens on Get Started, which detects your hardware and lists current models that fit it. Pick a chat model and an image model.
- Download them. Progress shows under Downloads in the logo menu.
- Run templates that use those models. They work offline.
Option 2: Start with Cloud Providers
- Get an API key from a provider:
- In NodeTool, go to Settings → Models & Providers
- Paste your API key
- Select the provider when using AI nodes
Detailed Guides
General
- Models Manager – Download and manage AI models
- Getting Started – First workflow
Local AI
- Supported Models – List of local models (llama.cpp, MLX, Whisper, Flux)
Cloud AI
- Providers Guide – Set up OpenAI, Anthropic, Google
- HuggingFace Integration – Access 500,000+ models
Advanced
- ComfyUI – Run an exported ComfyUI workflow as a node, locally, on a GPU worker, or on Comfy Cloud
- Self-Hosted Deployment – Secure deployments
- Deployment Guide – Cloud infrastructure