pricing.json

{
  "_about": "Approximate USD per 1M tokens, for the $ columns only. Update to your real rates. Matched by first substring hit; falls back to _default.",
  "_cache_about": "cache_write / cache_read are multipliers on that row's input rate. Anthropic bills a 5-minute cache write at 1.25x input and a cache read at 0.1x. OpenAI charges nothing extra to create a cache entry and bills cached input at 0.1x. Rows omitting either field inherit _default's.",
  "_default": { "in": 3.0, "out": 15.0, "cache_write": 1.25, "cache_read": 0.1 },
  "_order_note": "Matched by FIRST substring hit, so specific ids must precede generic ones — 'claude-opus-5' also contains 'opus'. Rates verified 2026-07-30 against the current model table.",
  "rates": [
    { "match": "gpt-5.5", "in": 2.5, "out": 15.0, "cache_write": 1.0, "cache_read": 0.1 },
    { "match": "gpt-5.4", "in": 2.5, "out": 15.0, "cache_write": 1.0, "cache_read": 0.1 },
    { "match": "fable",  "in": 10.0, "out": 50.0 },
    { "match": "mythos", "in": 10.0, "out": 50.0 },
    { "match": "opus-5", "in": 5.0,  "out": 25.0 },
    { "match": "opus-4-8", "in": 5.0, "out": 25.0 },
    { "match": "opus-4-7", "in": 5.0, "out": 25.0 },
    { "match": "opus-4-6", "in": 5.0, "out": 25.0 },
    { "match": "opus",   "in": 15.0, "out": 75.0 },
    { "match": "sonnet-5", "in": 3.0, "out": 15.0 },
    { "match": "sonnet", "in": 3.0,  "out": 15.0 },
    { "match": "haiku",  "in": 1.0,  "out": 5.0 }
  ]
}