{
  "source": "https://cutmyaispend.com",
  "description": "Per-provider AI cost-cutting playbooks (OpenAI, Anthropic, Bedrock, Azure OpenAI, Gemini).",
  "updated": "2026-08-06",
  "providers": [
    {
      "slug": "openai",
      "name": "OpenAI",
      "title": "How to cut your OpenAI API costs",
      "intro": "OpenAI spend concentrates in three places: over-provisioned models (GPT-5-class where mini-class would do), uncached repeated prefixes, and synchronous calls for offline work. The fixes, in order of leverage:",
      "plays": [
        {
          "method": "prompt-caching",
          "note": "Automatic 50% discount on repeated prefixes ≥1024 tokens — restructure prompts stable-first to maximize hits. No code flag needed."
        },
        {
          "method": "model-routing",
          "note": "Route classification/extraction to the mini tier; keep the frontier tier for reasoning-heavy requests. Most teams find 60–80% of traffic downgrades cleanly."
        },
        {
          "method": "batch-apis",
          "note": "The Batch API is a flat 50% off for anything that tolerates async completion."
        },
        {
          "method": "output-length-control",
          "note": "Structured Outputs + deliberate max_tokens caps cut premium-priced output tokens 20–60%."
        },
        {
          "method": "fix-the-context-layer",
          "note": "If agents re-read the same docs/emails each run, a memory layer (e.g. Mitosis Cortex) removes that spend entirely instead of discounting it."
        }
      ],
      "tip": "Check the usage dashboard by API key and model first — one over-provisioned endpoint is usually most of the bill."
    },
    {
      "slug": "anthropic-claude",
      "name": "Anthropic (Claude)",
      "title": "How to cut your Claude API costs",
      "intro": "Claude offers the deepest caching discount of any major provider — up to 90% off cached input — so prompt structure matters more here than anywhere else. The playbook:",
      "plays": [
        {
          "method": "prompt-caching",
          "note": "Explicit cache_control breakpoints give up to 90% off cached tokens (5-min and 1-hour TTLs). Put system prompt, tools, and reference docs above the breakpoint, byte-identical every call."
        },
        {
          "method": "model-routing",
          "note": "Haiku handles classification, extraction, and routine drafting at a fraction of Sonnet/Opus pricing; cascade upward only on failure."
        },
        {
          "method": "batch-apis",
          "note": "The Message Batches API is 50% off and stacks with prompt caching."
        },
        {
          "method": "context-hygiene",
          "note": "Long agent conversations dominate Claude bills — summarize history instead of replaying it, and trim tool outputs."
        },
        {
          "method": "fix-the-context-layer",
          "note": "Agent workloads that re-ingest context every session are the #1 Claude overspend pattern; persistent memory cuts it at the source."
        }
      ],
      "tip": "Cache-aware prompt ordering is worth more on Claude than any other single change — audit that first."
    },
    {
      "slug": "aws-bedrock",
      "name": "AWS Bedrock",
      "title": "How to cut your AWS Bedrock costs",
      "intro": "Bedrock spend hides inside the AWS bill, which makes attribution the first problem and model choice the second. In order:",
      "plays": [
        {
          "method": "cost-attribution-finops",
          "note": "Enable model-invocation logging and tag by application inference profile — untagged Bedrock spend is invisible in Cost Explorer."
        },
        {
          "method": "model-routing",
          "note": "Nova Micro/Lite are among the cheapest capable models anywhere; route simple tasks there before touching Claude-on-Bedrock pricing."
        },
        {
          "method": "prompt-caching",
          "note": "Bedrock supports prompt caching for Claude and Nova models — same stable-prefix discipline applies."
        },
        {
          "method": "batch-apis",
          "note": "Bedrock batch inference is ~50% off for async jobs."
        },
        {
          "method": "cheaper-and-open-models",
          "note": "Provisioned throughput only pays off at sustained high volume — price it against on-demand honestly before committing."
        }
      ],
      "tip": "Watch account-level quotas (RPM throttles) — teams often upgrade to bigger models to “fix” throttling that a quota request would solve for free."
    },
    {
      "slug": "azure-openai",
      "name": "Azure OpenAI",
      "title": "How to cut your Azure OpenAI costs",
      "intro": "Azure OpenAI adds a deployment-and-quota layer on top of OpenAI pricing, which creates its own waste (idle provisioned throughput) and its own levers:",
      "plays": [
        {
          "method": "cost-attribution-finops",
          "note": "Split deployments per team/app and tag resources — the single shared deployment pattern makes chargeback impossible."
        },
        {
          "method": "model-routing",
          "note": "Mini-class deployments for high-volume simple tasks; PTU (provisioned) capacity only for steady, latency-critical load."
        },
        {
          "method": "prompt-caching",
          "note": "Prefix caching discounts apply on Azure too — same stable-first prompt structure."
        },
        {
          "method": "batch-apis",
          "note": "Azure’s Global Batch is ~50% off for async workloads."
        },
        {
          "method": "output-length-control",
          "note": "Structured outputs + max_tokens discipline — output tokens carry the same 3–8× premium here."
        }
      ],
      "tip": "Audit PTU utilization monthly: idle provisioned throughput is the most common pure-waste line on Azure OpenAI bills."
    },
    {
      "slug": "google-gemini",
      "name": "Google (Gemini)",
      "title": "How to cut your Gemini API costs",
      "intro": "Gemini pricing is aggressive at the Flash tier, and its huge context windows tempt teams into expensive context stuffing. The levers:",
      "plays": [
        {
          "method": "model-routing",
          "note": "Flash and Flash-Lite handle most non-reasoning tasks at some of the lowest per-token prices on the market; reserve Pro for hard reasoning."
        },
        {
          "method": "context-hygiene",
          "note": "A 1M-token window is not a license to fill it — retrieval + reranking beats context stuffing on both cost and accuracy."
        },
        {
          "method": "prompt-caching",
          "note": "Context caching bills cached tokens at a steep discount for repeated large contexts (docs, codebases, videos)."
        },
        {
          "method": "batch-apis",
          "note": "Batch mode is 50% off for async jobs."
        },
        {
          "method": "fix-the-context-layer",
          "note": "If you keep re-uploading the same corpus per session, a persistent memory layer replaces the re-ingestion loop entirely."
        }
      ],
      "tip": "Giant-context convenience is the Gemini-specific overspend trap — measure tokens per task, not just price per token."
    }
  ]
}