{
  "lastUpdated": "2026-06-04T00:58:02Z",
  "schemaVersion": 2,
  "tokensPerRound": { "input": 5000, "output": 2000 },
  "providers": [
    {
      "id": "gemini-free",
      "name": "Gemini (Google) — free tier",
      "shortName": "Gemini Flash (free tier)",
      "model": "gemini-2.5-flash",
      "inputPerM": 0.00,
      "outputPerM": 0.00,
      "contextWindow": "1M",
      "maxOutput": "8K",
      "estPerRound": 0.000,
      "estNote": "free while AI Studio billing is disabled",
      "billingUrl": "https://aistudio.google.com/apikey",
      "tier1Rpm": "15",
      "tier1Tpm": "—",
      "freeTier": "1500 RPD",
      "rateLimitNotes": "Free while billing disabled. Daily quota resets at midnight Pacific.",
      "recommendationTag": "cheapest",
      "recommendationNote": "Free while AI Studio billing is disabled. 1M-token context window, 1500 requests/day quota, synthesizes Builder output reliably. Catch: enable billing on AI Studio and routing can quietly flip to paid tier on the same model name."
    },
    {
      "id": "gemini-paid",
      "name": "Gemini (Google) — paid",
      "shortName": "Gemini Flash (paid)",
      "model": "gemini-2.5-flash",
      "inputPerM": 0.075,
      "outputPerM": 0.30,
      "contextWindow": "1M",
      "maxOutput": "8K",
      "estPerRound": 0.001,
      "billingUrl": "https://aistudio.google.com/apikey",
      "tier1Rpm": "2K",
      "tier1Tpm": "4M",
      "freeTier": "n/a",
      "rateLimitNotes": "Paid-tier limits are very generous; cost discipline matters more than RPM.",
      "recommendationTag": null,
      "recommendationNote": null
    },
    {
      "id": "grok",
      "name": "Grok (xAI)",
      "shortName": "Grok",
      "model": "grok-4-fast-non-reasoning",
      "inputPerM": 0.20,
      "outputPerM": 0.50,
      "contextWindow": "256K",
      "maxOutput": "8K",
      "estPerRound": 0.002,
      "billingUrl": "https://console.x.ai/billing",
      "tier1Rpm": "~60",
      "tier1Tpm": "varies",
      "freeTier": "Limited",
      "rateLimitNotes": "API availability and quotas vary; check console for current tier.",
      "recommendationTag": null,
      "recommendationNote": null
    },
    {
      "id": "deepseek",
      "name": "DeepSeek",
      "shortName": "DeepSeek",
      "model": "deepseek-chat",
      "inputPerM": 0.27,
      "outputPerM": 1.10,
      "contextWindow": "64K",
      "maxOutput": "8K",
      "estPerRound": 0.004,
      "billingUrl": "https://platform.deepseek.com/top_up",
      "tier1Rpm": "~60",
      "tier1Tpm": "varies",
      "freeTier": "None",
      "rateLimitNotes": "Cheapest paid option, but consistently the slowest responder in the hive (60–90s/round).",
      "recommendationTag": "cheapest",
      "recommendationNote": "Cheapest reliable paid Builder — roughly 10x cheaper per token than Claude or ChatGPT. Caveat: consistently the slowest responder in the hive, often 60–90 seconds per round. If you don't mind the wait, it's the best value."
    },
    {
      "id": "together",
      "name": "Together AI",
      "shortName": "Together AI",
      "model": "Llama-3.3-70B-Instruct-Turbo",
      "inputPerM": 0.88,
      "outputPerM": 0.88,
      "contextWindow": "128K",
      "maxOutput": "4K",
      "estPerRound": 0.006,
      "billingUrl": "https://api.together.ai/settings/billing",
      "tier1Rpm": "~600",
      "tier1Tpm": "varies",
      "freeTier": "Trial credits",
      "rateLimitNotes": "Open-weight model gateway; generous limits, not yet hive-tested in production.",
      "recommendationTag": null,
      "recommendationNote": null
    },
    {
      "id": "mistral",
      "name": "Mistral",
      "shortName": "Mistral",
      "model": "mistral-large-latest",
      "inputPerM": 2.00,
      "outputPerM": 6.00,
      "contextWindow": "128K",
      "maxOutput": "8K",
      "estPerRound": 0.022,
      "billingUrl": "https://console.mistral.ai/api-keys",
      "tier1Rpm": "~30",
      "tier1Tpm": "varies",
      "freeTier": "Trial credits",
      "rateLimitNotes": "Low RPS on paid Tier 1. A parallel 6-bee hive can trigger 429 even with token budget remaining. Contact Mistral support for tier raise.",
      "recommendationTag": "balanced",
      "recommendationNote": "Similar territory to ChatGPT — fast, reliable, with a distinct European model lineage that adds genuine diversity to the hive. Watch the low Tier 1 RPM on paid accounts."
    },
    {
      "id": "chatgpt",
      "name": "ChatGPT (OpenAI)",
      "shortName": "ChatGPT",
      "model": "gpt-4.1",
      "inputPerM": 2.00,
      "outputPerM": 8.00,
      "contextWindow": "1M",
      "maxOutput": "32K",
      "estPerRound": 0.026,
      "billingUrl": "https://platform.openai.com/settings/organization/billing/overview",
      "tier1Rpm": "500",
      "tier1Tpm": "30K",
      "freeTier": "None",
      "rateLimitNotes": "Tiers scale with usage history; $5 credit unlocks most user needs.",
      "recommendationTag": "balanced",
      "recommendationNote": "Strong all-rounder — fast, reliable formatting compliance, consistent convergence behavior. A good default for long-form documents where Builder speed matters."
    },
    {
      "id": "cohere",
      "name": "Cohere",
      "shortName": "Cohere",
      "model": "command-r-plus",
      "inputPerM": 2.50,
      "outputPerM": 10.00,
      "contextWindow": "128K",
      "maxOutput": "4K",
      "estPerRound": 0.033,
      "billingUrl": "https://dashboard.cohere.com/billing",
      "tier1Rpm": "~100",
      "tier1Tpm": "varies",
      "freeTier": "Trial credits",
      "rateLimitNotes": "Generous trial credits; not yet hive-tested in production.",
      "recommendationTag": null,
      "recommendationNote": null
    },
    {
      "id": "claude",
      "name": "Claude (Anthropic)",
      "shortName": "Claude",
      "model": "claude-sonnet-4-6",
      "inputPerM": 3.00,
      "outputPerM": 15.00,
      "contextWindow": "200K",
      "maxOutput": "8K",
      "estPerRound": 0.045,
      "billingUrl": "https://console.anthropic.com/settings/billing",
      "tier1Rpm": "50",
      "tier1Tpm": "50K",
      "freeTier": "None",
      "rateLimitNotes": "Tiers progress automatically with paid spend over time.",
      "recommendationTag": "highest-capability",
      "recommendationNote": "Frequently the best at nuanced voice, precise instruction-following, and detailed reasoning on complex documents. For high-stakes documents (RFP responses, executive summaries, board memos), the premium pays off in fewer rounds to converge."
    },
    {
      "id": "perplexity",
      "name": "Perplexity",
      "shortName": "Perplexity",
      "model": "sonar-pro",
      "inputPerM": 3.00,
      "outputPerM": 15.00,
      "contextWindow": "200K",
      "maxOutput": "8K",
      "estPerRound": 0.045,
      "billingUrl": "https://console.perplexity.ai",
      "tier1Rpm": "~50",
      "tier1Tpm": "varies",
      "freeTier": "None",
      "rateLimitNotes": "$5/month recurring subscription tier (enable auto-pay at signup) covers most usage; $50/month otherwise.",
      "recommendationTag": null,
      "recommendationNote": null
    }
  ]
}