Haz que tu router sea listado y medido
TKX clasifica el mercado de routers por medición, no por afirmaciones. Implementa el estándar abierto — la mayoría está a 30 minutos — y nuestro colector horario te verifica automáticamente. Tú expones, nosotros medimos.
The four levels
GET /v1/models. TKX measures your model coverage every hour. Evidence chip: Coverage measured.L2 pricing formats (any of these four)
Prices must be machine-readable per model. We already auto-detect all four dialects in production:
{"id":"org/model","pricing":{"prompt":"0.0000025","completion":"0.00001"}} — USD per token (strings).{"id":"org/model","input_price":2.5e-6,"output_price":1e-5} — USD per token (numbers). Used by Requesty.{"id":"org/model","input_token_price_per_m":25000,"output_token_price_per_m":100000} — 1/10000 USD per 1M tokens. Used by Novita.{"model_name":"model","quota_type":0,"model_ratio":0.125,"completion_ratio":8} — ratio × $2/1M input, × completion_ratio for output; your existing /api/pricing works as-is. quota_type 1 (per-call) rows are skipped.Recommended for new implementations: dialect A (OpenRouter-compatible). Running one-api/new-api? Your public /api/pricing is already dialect D — just submit it. Sanity valves apply: fewer than 3 priced models, prices outside $0.0005–$2000 /1M, or a median markup outside −80%…+300% vs official list will reject the source — garbage can't buy a ranking.
Why implement it
<a href="https://tkx.org/a/your-slug/">Measured on TKX</a> to your site. Your TKX profile page links back to you.Submit
Email echo@minax.io with your router name + base URL. If your /v1/models is already public, say so — verification is automatic within the hour. Machine contract: openapi.json.