Data Sources

Every number on AIpricly has an upstream source. Here they all are.

Aggregators

• LiteLLM — community-maintained pricing JSON we sync daily for the bulk of model entries.
• Artificial Analysis — quality benchmark scores (Intelligence Index).
• LMArena — human-preference Elo, used in model detail pages where applicable.

Provider pricing pages

• Alibaba: qwenlm.com
• Anthropic: anthropic.com
• AssemblyAI: assemblyai.com
• Black Forest Labs: blackforestlabs.ai
• Cartesia: cartesia.ai
• Cohere: cohere.com
• Deepgram: deepgram.com
• DeepSeek: deepseek.com
• ElevenLabs: elevenlabs.io
• Google: ai.google.dev
• Hailuo: hailuoai.video
• Hume: hume.ai
• Ideogram: ideogram.ai
• Jina AI: jina.ai
• Kling: klingai.com
• Luma: lumalabs.ai
• Meta: ai.meta.com
• Midjourney: midjourney.com
• Mistral: mistral.ai
• Mixedbread: mixedbread.ai
• Moonshot: moonshot.cn
• Nomic: nomic.ai
• OpenAI: openai.com
• Pika: pika.art
• Play.ht: play.ht
• Recraft: recraft.ai
• Runway: runwayml.com
• Stability: stability.ai
• Tencent: hunyuan.tencent.com
• Voyage: voyageai.com
• xAI: x.ai

Open data

Our entire data layer (the data/ directory, including augments and benchmark snapshots) is version-controlled in our public repository. You can audit any number against its commit history.

Public API

The same normalized model data is exposed as a read-only JSON API. No auth, CORS open, daily-refreshed, free under CC-BY-4.0. Please credit “AIpricly” and link back when you display the data.

Self-describing entry point: /api/v1 — hit it once and discover every endpoint.

• GET /api/v1/meta — registry counts and last sync timestamp.
• GET /api/v1/models — list. Optional query params:
  • ?modality=text | image | video | audio | embedding | rerank
  • ?provider=<slug> — e.g. openai, anthropic
  • ?capability=vision | function-calling | reasoning | …
  • ?limit=N — positive integer, max 500
• GET /api/v1/models/{provider}/{model} — single model by canonical id.

Responses are cached at the CDN for one hour (s-maxage=3600) with stale-while-revalidate. If you ingest at scale, mirror /api/v1/models daily rather than fan-out from end users.

Embed widget

Drop a price card for any model into your blog, docs, or README with a single <script> tag. The widget self-updates daily, renders inside an iframe (no style conflicts with your page), and emits a sibling <a> attribution link in the parent DOM — so embedding our widget adds a real backlink to your site, not buried inside an iframe Google can't crawl.

<script src="https://aipricly.com/embed.js"
        data-model="anthropic/claude-haiku-4-5"
        data-theme="auto"
        data-width="320"></script>

• data-model (required) — the canonical model id from /api/v1/models, e.g. openai/gpt-5
• data-theme — light (default), dark, or auto (follows prefers-color-scheme)
• data-width — px integer 220–720, default 320

Misconfigured embeds fail silently with a console warning — they will not break your page layout.