AI API Catalog

Compare prices, providers, and capabilities for 2,000+ AI models available via API — from LLMs to video generation.

Live Catalog · Report an Issue · Add a Model

---

Why AI API Catalog?

• One place, all providers — LiteLLM, OpenRouter, fal.ai, Replicate, Fireworks, Runway, kie.ai, piapi.ai, and 17 more aggregated into a single searchable catalog
• Real pricing — per-token, per-second, per-image, per-video with best-price highlighting across providers
• Beyond LLMs — video generation, image generation, TTS, STT, music generation, embeddings all in one place
• Weekly auto-updates — scripts pull fresh data from provider APIs every Monday
• Community-maintained — missing a model or wrong price? Open a PR

---

What's in the catalog

Category	Examples	Models
💬 LLM / Chat	GPT-4o, Claude 4, Gemini 2.5 Pro, DeepSeek R1	800+
🎨 Image generation	FLUX.1, Stable Diffusion, DALL-E 3, Midjourney	500+
🎬 Video generation	Veo 3.1, Kling 2.1, Sora 2, WAN 2.1, Seedance	200+
🔊 TTS	ElevenLabs, OpenAI TTS, Kokoro, Cartesia	100+
🎤 STT	Whisper, Deepgram, AssemblyAI	30+
🎵 Music	Suno, Udio, MusicGen	20+
📊 Embeddings	text-embedding-3, Voyage-4, Cohere Embed	50+

25+ providers including: OpenAI, Anthropic, Google, Mistral, Groq, DeepSeek, OpenRouter, Together AI, Fireworks AI, Replicate, fal.ai, HuggingFace, Runway, Stability AI, ElevenLabs, Deepgram, piapi.ai, wavespeed.ai, kie.ai, aimlapi.com

---

Quick Start

No build step. No dependencies. Open directly in browser:

git clone https://github.com/jurczykpawel/ai-api-catalog.git
cd ai-api-catalog
python3 -m http.server 8080
# Open http://localhost:8080

Note: Must be served via HTTP, not opened as file:// (fetch() calls won't work).

---

Updating data

Run daily to pull fresh pricing from all provider APIs:

bash scripts/update-litellm.sh

This runs 9 steps: fetches LiteLLM pricing, OpenRouter (640+ models via frontend API), fal.ai (700+ models), HuggingFace, AIMLAPI, piapi.ai, WaveSpeed, kie.ai, Runway — then merges and validates.

Optional env vars:

export HF_TOKEN=hf_xxx              # Higher HuggingFace rate limits
export REPLICATE_API_TOKEN=r8_xxx   # Enable Replicate fetch (450+ models)
export FIREWORKS_API_KEY=fw_xxx     # Enable Fireworks fetch (250 models)
export RUNWAY_API_KEY=key_xxx       # Enable Runway fetch
export OR_NOTIFY_WEBHOOK=https://...  # Discord/Slack alert if OpenRouter API changes

After updating:

git add data/ && git commit -m "chore: update catalog $(date +%Y-%m-%d)"

---

Project structure

ai-api-catalog/
├── index.html                  # Single-file frontend (Tailwind CDN + vanilla JS)
├── data/
│   ├── models.json             # Generated — 2000+ models with pricing (do not edit)
│   ├── models-manual.json      # Hand-curated models: video, image, audio, niche LLMs
│   ├── providers.json          # Provider metadata: name, logo color, type
│   ├── categories.json         # Categories with icons and EN/PL translations
│   ├── provider-patches.json   # Manual provider additions for specific model IDs
│   └── *-raw.json              # Raw API responses (auto-fetched, do not edit)
├── scripts/
│   ├── update-litellm.sh       # Main update script — run this daily
│   ├── merge-data.py           # Merges all sources into models.json
│   ├── validate-schema.py      # JSON schema validation before deploy
│   ├── model-radar.py          # Detects new models vs last snapshot
│   ├── fetch-openrouter.py     # OpenRouter (640+ models, frontend + v1 fallback)
│   ├── fetch-fal.py            # fal.ai (700+ models)
│   ├── fetch-huggingface.py    # HuggingFace Inference API
│   ├── fetch-aimlapi.py        # AIMLAPI
│   ├── fetch-replicate.py      # Replicate (requires token)
│   ├── fetch-fireworks.py      # Fireworks AI (requires key)
│   ├── fetch-piapi.py          # piapi.ai curated pricing
│   ├── fetch-wavespeed.py      # WaveSpeed AI curated pricing
│   ├── fetch-kie.py            # kie.ai curated pricing
│   └── fetch-runway.py         # Runway ML curated pricing
└── n8n/
    ├── model-radar-workflow.json   # Weekly update + change detection
    └── news-radar-workflow.json    # Daily AI news → extract new models → Discord

---

Data schema

Model entry (data/models-manual.json — source of truth for manual entries)

{
  "id": "veo-3-1",
  "name": "Veo 3.1",
  "category": "video_generation",
  "description": "Google video model with synchronized native audio.",
  "tags": ["text-to-video", "google", "audio-sync"],
  "capabilities": ["text_to_video", "audio_generation"],
  "open_source": false,
  "local_available": false,
  "providers": [
    {
      "provider_id": "google",
      "pricing": { "per_second": 0.40, "notes": "720p/1080p via Gemini API" },
      "url": "https://ai.google.dev/gemini-api/docs/video",
      "affiliate_url": null,
      "available": true
    }
  ],
  "updated_at": "2026-03-10",
  "source": "manual"
}

Categories: llm · image_generation · video_generation · audio_tts · audio_stt · music_generation · embedding · moderation

Capabilities: vision · reasoning · function_calling · web_search · prompt_caching · text_to_video · image_to_video · image_editing · audio_generation · json_mode

Pricing units: input_per_1m · output_per_1m · per_image · per_second · per_video · per_minute · per_song · notes

---

Contributing

Contributions are welcome — especially pricing corrections and missing models.

Fix a price or add a provider to an existing model

Edit data/models-manual.json. Find the model by id and add/update the provider entry. Then validate:

python3 scripts/validate-schema.py data/models.json

Add a new model (video, image, audio, niche LLM)

1. Add an entry to data/models-manual.json following the schema above
2. Run bash scripts/update-litellm.sh to regenerate models.json
3. Run python3 scripts/validate-schema.py data/models.json — must return zero errors
4. Open a PR with a brief description of the model and source for the pricing

Add a new provider (automated fetch)

1. Add provider metadata to data/providers.json
2. Create scripts/fetch-PROVIDER.py (use fetch-piapi.py as a template for curated providers, or fetch-fal.py for API-based fetch)
3. Add a step to scripts/update-litellm.sh
4. Add --provider argument and parser in scripts/merge-data.py (use parse_curated())

Provider patches (manual additions without a full fetch script)

For models that exist in LiteLLM but are missing a provider entry, use data/provider-patches.json:

{
  "patches": [
    {
      "model_id": "some-model",
      "provider": {
        "provider_id": "fireworks",
        "pricing": { "per_second": 0.05 },
        "url": "https://fireworks.ai/models/fireworks/some-model",
        "available": true
      }
    }
  ]
}

---

Tech stack

• Frontend: vanilla JavaScript, Tailwind CSS (CDN), Syne + DM Sans + JetBrains Mono
• Data pipeline: Python 3 scripts (stdlib only — no pip install required)
• Primary data sources: LiteLLM model prices, OpenRouter API, fal.ai API
• Automation: GitHub Actions for daily updates and news radar pipeline
• Hosting: nginx static file serving

---

Roadmap

• 2000+ models across 8 categories
• 25+ providers with live pricing
• Weekly automated updates via shell scripts
• OpenRouter frontend API (640 models including image gen)
• Fireworks AI as full data source
• Free API detection (free_type: open vs tier)
• OSS/local availability auto-detection
• Model Radar — weekly diff and change detection
• AI News Radar — daily RSS → Claude → Discord
• Price change history (track over time)
• Affiliate links (OpenRouter, Replicate, fal.ai, ElevenLabs, piapi.ai)
• Email alerts for price changes
• B2B catalog API (programmatic access)

---

Deployment

rsync -avz --exclude='.git' . mikrus:~/sites/ai-api-catalog/

nginx config:

server {
    root /home/ubuntu/sites/ai-api-catalog;
    index index.html;
    location / { try_files $uri $uri/ =404; }
}

---

Acknowledgments

• LiteLLM — the primary source for LLM pricing data
• OpenRouter — 640+ models including multimodal
• TechSkills Academy — maintainer

---

Support

The catalog is free and open source, but model prices change weekly and the crawler doesn't run on air. You can top up the tank:

License

MIT — see LICENSE