{
  "version": 1,
  "lastUpdated": "2026-07-22",
  "source": "https://openviglet.github.io/model-catalog",
  "modality": "audio",
  "vendors": {
    "gemini": [
      {
        "id": "gemini-2.0-flash",
        "label": "Gemini 2.0 Flash",
        "kind": "CHAT",
        "contextWindow": 1048576,
        "maxOutputTokens": 8192,
        "capabilities": [
          "vision",
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text",
            "video"
          ],
          "output": [
            "image",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.1,
          "outputPer1M": 0.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "benchmarks": {
          "intelligenceIndex": 12.3,
          "scores": {
            "math": {
              "value": 21.7
            }
          },
          "indicative": true,
          "note": "Cited third-party benchmark — verify at the source.",
          "source": "Artificial Analysis",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "artificial-analysis",
          "committed",
          "litellm"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "gemini"
      },
      {
        "id": "gemini-2.0-flash-lite",
        "label": "Gemini 2.0 Flash-Lite",
        "kind": "CHAT",
        "contextWindow": 1048576,
        "maxOutputTokens": 8192,
        "capabilities": [
          "vision",
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text",
            "video"
          ],
          "output": [
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.075,
          "outputPer1M": 0.3,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "benchmarks": {
          "intelligenceIndex": 8.8,
          "indicative": true,
          "note": "Cited third-party benchmark — verify at the source.",
          "source": "Artificial Analysis",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "artificial-analysis",
          "committed",
          "litellm"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "gemini"
      },
      {
        "id": "gemini-2.5-flash",
        "label": "Gemini 2.5 Flash",
        "kind": "CHAT",
        "contextWindow": 1048576,
        "maxOutputTokens": 65535,
        "capabilities": [
          "vision",
          "tools",
          "reasoning"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "pdf",
            "text",
            "video"
          ],
          "output": [
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.3,
          "outputPer1M": 2.5,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "benchmarks": {
          "intelligenceIndex": 14.1,
          "scores": {
            "math": {
              "value": 60.3
            }
          },
          "indicative": true,
          "note": "Cited third-party benchmark — verify at the source.",
          "source": "Artificial Analysis",
          "lastVerified": "2026-07-22"
        },
        "performance": {
          "throughputTps": 227.181,
          "latencyTtftSec": 0.373,
          "indicative": true,
          "note": "Cited third-party benchmark — verify at the source.",
          "source": "Artificial Analysis",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "artificial-analysis",
          "committed",
          "litellm"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "gemini"
      },
      {
        "id": "gemini-2.5-pro",
        "label": "Gemini 2.5 Pro",
        "kind": "CHAT",
        "contextWindow": 1048576,
        "maxOutputTokens": 65535,
        "capabilities": [
          "vision",
          "tools",
          "reasoning"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "pdf",
            "text",
            "video"
          ],
          "output": [
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 1.25,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "benchmarks": {
          "intelligenceIndex": 25.8,
          "scores": {
            "coding": {
              "value": 33.3
            },
            "math": {
              "value": 87.7
            }
          },
          "indicative": true,
          "note": "Cited third-party benchmark — verify at the source.",
          "source": "Artificial Analysis",
          "lastVerified": "2026-07-22"
        },
        "performance": {
          "throughputTps": 136.211,
          "latencyTtftSec": 20.35,
          "indicative": true,
          "note": "Cited third-party benchmark — verify at the source.",
          "source": "Artificial Analysis",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "artificial-analysis",
          "committed",
          "litellm"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "gemini"
      }
    ],
    "openai": [
      {
        "id": "gpt-audio",
        "label": "gpt-audio",
        "kind": "CHAT",
        "contextWindow": 128000,
        "maxOutputTokens": 16384,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 2.5,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-whisper",
        "label": "gpt-realtime-whisper",
        "kind": "TRANSCRIPTION",
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "text"
          ]
        },
        "sources": [
          "committed",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-4o-mini-tts",
        "label": "gpt-4o-mini-tts",
        "kind": "SPEECH",
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio"
          ]
        },
        "pricing": {
          "inputPer1M": 2.5,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-4o-mini-tts-2025-03-20",
        "label": "gpt-4o-mini-tts-2025-03-20",
        "kind": "SPEECH",
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio"
          ]
        },
        "pricing": {
          "inputPer1M": 2.5,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-4o-mini-tts-2025-12-15",
        "label": "gpt-4o-mini-tts-2025-12-15",
        "kind": "SPEECH",
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio"
          ]
        },
        "pricing": {
          "inputPer1M": 2.5,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-audio-1.5",
        "label": "gpt-audio-1.5",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 16384,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 2.5,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-audio-2025-08-28",
        "label": "gpt-audio-2025-08-28",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 16384,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 2.5,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-audio-mini",
        "label": "gpt-audio-mini",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 16384,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.6,
          "outputPer1M": 2.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-audio-mini-2025-10-06",
        "label": "gpt-audio-mini-2025-10-06",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 16384,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.6,
          "outputPer1M": 2.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-audio-mini-2025-12-15",
        "label": "gpt-audio-mini-2025-12-15",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 16384,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.6,
          "outputPer1M": 2.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime",
        "label": "gpt-realtime",
        "kind": "SPEECH",
        "contextWindow": 32000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 4,
          "outputPer1M": 16,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-1.5",
        "label": "gpt-realtime-1.5",
        "kind": "SPEECH",
        "contextWindow": 32000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 4,
          "outputPer1M": 16,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-2",
        "label": "gpt-realtime-2",
        "kind": "SPEECH",
        "contextWindow": 32000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 4,
          "outputPer1M": 16,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-2.1",
        "label": "gpt-realtime-2.1",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 32000,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 4,
          "outputPer1M": 24,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-2.1-mini",
        "label": "gpt-realtime-2.1-mini",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.6,
          "outputPer1M": 2.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-2025-08-28",
        "label": "gpt-realtime-2025-08-28",
        "kind": "SPEECH",
        "contextWindow": 32000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 4,
          "outputPer1M": 16,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-mini",
        "label": "gpt-realtime-mini",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.6,
          "outputPer1M": 2.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-mini-2025-10-06",
        "label": "gpt-realtime-mini-2025-10-06",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.6,
          "outputPer1M": 2.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      },
      {
        "id": "gpt-realtime-mini-2025-12-15",
        "label": "gpt-realtime-mini-2025-12-15",
        "kind": "SPEECH",
        "contextWindow": 128000,
        "maxOutputTokens": 4096,
        "capabilities": [
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text"
          ],
          "output": [
            "audio",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.6,
          "outputPer1M": 2.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm",
          "openai-api"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "openai"
      }
    ],
    "vertex-ai": [
      {
        "id": "gemini-2.0-flash",
        "label": "Gemini 2.0 Flash",
        "kind": "CHAT",
        "contextWindow": 1048576,
        "maxOutputTokens": 8192,
        "capabilities": [
          "vision",
          "tools"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "text",
            "video"
          ],
          "output": [
            "image",
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.1,
          "outputPer1M": 0.4,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "vertex-ai"
      },
      {
        "id": "gemini-2.5-flash",
        "label": "Gemini 2.5 Flash",
        "kind": "CHAT",
        "contextWindow": 1048576,
        "maxOutputTokens": 65535,
        "capabilities": [
          "vision",
          "tools",
          "reasoning"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "pdf",
            "text",
            "video"
          ],
          "output": [
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 0.3,
          "outputPer1M": 2.5,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "vertex-ai"
      },
      {
        "id": "gemini-2.5-pro",
        "label": "Gemini 2.5 Pro",
        "kind": "CHAT",
        "contextWindow": 1048576,
        "maxOutputTokens": 65535,
        "capabilities": [
          "vision",
          "tools",
          "reasoning"
        ],
        "modalities": {
          "input": [
            "audio",
            "image",
            "pdf",
            "text",
            "video"
          ],
          "output": [
            "text"
          ]
        },
        "pricing": {
          "inputPer1M": 1.25,
          "outputPer1M": 10,
          "currency": "USD",
          "unit": "per_1M_tokens",
          "indicative": true,
          "note": "Indicative US list price — verify with the vendor.",
          "source": "litellm",
          "lastVerified": "2026-07-22"
        },
        "sources": [
          "committed",
          "litellm"
        ],
        "lastVerified": "2026-07-22",
        "vendor": "vertex-ai"
      }
    ]
  }
}