{
  "source_file": "0930",
  "source_platform": "Dev.to",
  "author": "Edward Izgorodin",
  "title_listed": "Rename a Tool, Lose the Entire Cache. The Vendor Documents It.",
  "keywords": "",
  "domain": "大模型算法",
  "content_type": "深度技术解析",
  "is_paid_flag": "否",
  "url": "https://dev.to/izgorodin/cache-reads-got-cheaper-this-week-why-does-your-agent-still-pay-full-price-1p3h",
  "publish_date": "2026-09-10",
  "corpus_stage": "",
  "notes": "",
  "id": "a8bcc1051de4",
  "collected_at": "2026-10-02T14:03:14",
  "title": "Rename a Tool, Lose the Entire Cache. The Vendor Documents It. - DEV Community",
  "body_markdown": "A cache discount does not reduce what a miss costs, it widens the gap: one Anthropic footnote now puts it at forty where the standard multiplier gives ten.\nFive vendors, pages opened on 2026-09-07\nThere is no industry rate for a cached read. Figures are per million tokens. Anthropic prints its two multipliers; the rest of that column is my division.\nProvider and model\nBase input\nCached read\nMultiplier\nProbed\nSource\nAnthropic, Claude Fable 5.1\n$10.00\n$0.25\n0.025x\n2026-09-07\npricing\nAnthropic, Claude Fable 5\n$10.00\n$1.00\n0.1x\n2026-09-07\npricing\nAnthropic, Claude Opus 5\n$5.00\n$0.50\n0.1x\n2026-09-07\npricing\nOpenAI, gpt-6-astra (standard, short context)\n$10.00\n$1.00\n0.1x\n2026-09-07\npricing\nGoogle, Gemini 3.8 Flash\n$0.75\n$0.075\n0.1x\n2026-09-07\npricing\nGroq, caching-enabled models\nnot on this page\n50 percent discount\n0.5x\n2026-09-07\ncaching docs\nDeepSeek, deepseek-v4-pro (off-peak)\n$0.66\n$0.022\n0.033x\n2026-09-07\npricing\nDeepSeek, deepseek-v4-flash (off-peak)\n$0.22\n$0.007\n0.032x\n2026-09-07\npricing\nThe change is one footnote on the Anthropic pricing page: \"Cache hits and refreshes on Claude Fable 5.1 and Claude Mythos 5.1 are priced at 0.025x the base input price. All other models use the standard 0.1x multiplier.\" The\nmodel overview page\nsays Fable 5.1 holds Fable 5 input and output prices, cache reads at a quarter.\nGroq documents the discount, not the base rates, and only on hits: 0.5x is a best case, a mixed run bills above it. Four multipliers on one table, 0.025x, 0.033x, 0.1x and 0.5x, is what no industry rate looks like in practice. Google charges cache storage on top, $0.50 per million tokens per hour through 2026-12-31 and $1.00 from 2027-01-01, dating the Gemini row too: $0.75 and $0.075 through 2026-12-31, $1.50 and $0.15 from 2027-01-01. The OpenAI row is the standard tier at short context; the same model in fast mode lists $20.00 base input and $2.00 cached. The DeepSeek rows are the off-peak half of a clock: the page states that off-peak rates are half of the peak rates, with peak hours 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, so both columns double at peak and the multiplier does not move. It also no longer prices one DeepSeek, which is why the model is named in each row.\nTwo posts priced the read\nOn 2026-09-03 max_quimby published\nFable 5.1's Real Story Is the 75% Cache Price Cut\n: the Fable 5 against Fable 5.1 price table moves on one row, cache reads falling from $1.00 to $0.25 per million tokens, the other rows unchanged.\nOn 2026-09-04 aidiveyt published\nFable 5.1 cut cache reads by 75%. Cost per task went up 20% anyway.\n: the pelican Simon Willison drew at every effort level cost 10 cents on low effort against $3.30 at maximum, with 65,927 output tokens, attributed there to the effort dial, not caching.\nIn\ncomment 3e615\non 2026-09-04, max_quimby named the operational half: \"The moment your orchestrator reorders tools, injects a timestamp, or rotates a system-prompt fragment, you bust the cache and quietly pay full input price on the re-read.\" He reports identical-looking loops differing three to four times in bill on prompt determinism alone, a production report, not a controlled experiment. Neither post prices the miss, which errors nowhere and shows only as a line item.\nThe discount raised the price of a mistake\nTake a system prefix of 20,000 tokens over 50 turns: 1,000,000 tokens a session. All hits on Fable 5.1, at $0.25 per million, cost $0.25.\nOne clock in the system prompt changes that: a timestamp that ticks per request alters an early token, the prefix stops matching, and every turn misses, the same 1,000,000 tokens at the $10.00 base input price, $10.00. Forty times the disciplined run, for one field.\nThe same two lines on Claude Opus 5, $5.00 base input and $0.50 cached: $0.50 against $5.00, a factor of ten.\nThe miss price did not move: Fable 5 and Fable 5.1 both bill base input at $10.00 per million. The ratio changed, and the ratio is what prefix discipline is worth per turn.\nCache writes bill separately, $12.50 per million for the five-minute window on the Anthropic pricing page, over base input, so a prefix rewritten every turn bills above the miss line, not below.\nWhat breaks the prefix, according to the vendor\nOn the\nprompt caching page\nthe cache follows a hierarchy of tools, then system, then messages: a change at one level invalidates that level and every one after.\nWhat changes\nTools cache\nSystem cache\nMessages cache\nTool definitions: names, descriptions, parameters\ninvalidated\ninvalidated\ninvalidated\nWeb search toggle\nvalid\ninvalidated\ninvalidated\nSpeed setting, fast against standard\nvalid\ninvalidated\ninvalidated\nTool choice parameter\nvalid\nvalid\ninvalidated\nImages added or removed\nvalid\nvalid\ninvalidated\nThinking parameters\nmodel specific\nmodel specific\ninvalidated\nEffort setting\nmodel specific\nmodel specific\ninvalidated\nThe first row is expensive: \"Modifying tool definitions (names, descriptions, parameters) invalidates the entire cache\". A registry sorting tools by dictionary iteration order pays it on every deploy, and nobody decided to.\nTwo rows carry an exception. Where per-message effort is supported, an effort change carried in a message with role system leaves the cached prefix intact. On Fable 5.1, Mythos 5.1, Fable 5, Mythos 5, Opus 4.8 and Opus 5, an instruction appended as a message with role system, rather than edited into the top-level system field, leaves the system and message caches alone. The page excludes Claude Sonnet 5 by name.\nThe hit rate is a meter, not a setting\nThe same page documents three usage fields: cache_read_input_tokens for what was served, cache_creation_input_tokens for what was written, input_tokens for what was neither, and total_input_tokens = cache_read_input_tokens + cache_creation_input_tokens + input_tokens, making input_tokens a remainder, not the size of what you sent.\nDeepSeek exposes the same split as prompt_cache_hit_tokens and prompt_cache_miss_tokens, and its\ncache guide\ndescribes the cache as best effort, no guarantee of a hit. Groq states the matching rule: \"Cache hits are only possible for exact prefix matches within a prompt.\"\nIn its\nengineering post of 2025-07-18\nthe Manus team argued that \"the KV-cache hit rate is the single most important metric for a production-stage AI agent\", reporting an average input-to-output token ratio around 100 to 1, a vendor self-report. Log it per prompt version: a drop that stays down is a prompt change nobody flagged.\nWorking memory and long-term memory are two different things\nEvery cache here has a stated lifetime: Anthropic prices cache writes in five-minute and one-hour windows, the DeepSeek guide a disk cache cleared over hours to days. That is working memory.\nThe durable layer has separate economics.\nMemGPT\n(Packer et al., 2023) framed it as an operating-system hierarchy: what the paper calls main context, analogous to RAM, against external context, analogous to disk. Prefix stability makes the fast tier cheap and decides nothing about what survives the session. Whether context engineering and memory engineering are one discipline is argued both ways: one survey treats memory as a component of context engineering (Mei et al., 2025), another separates them (Hu et al., December 2025), both cited on the\noriginal page\n.\nWhat this does not prove\nThis is a reading of documentation and price tables, not a measurement. I opened eight vendor pages on 2026-09-07. No agent was run, no bill reproduced, no hit rate recorded.\nThe arithmetic is a calculation on list prices assuming every turn rereads the full prefix, while real loops have partial matches, several breakpoints and expiry between turns. The cost per task figures and the three to four times spread belong to other authors.\nPrices are a snapshot, and one row carries its own expiry: the Gemini figures hold through 2026-12-31 by Google's note. The price table on the original page carries an earlier check date, so where the two disagree, trust the dated table above.\nThe invalidation table is one vendor's behavior and does not transfer. It says what would break a prefix, not that anything is breaking yours. The cheap way to find out is those three usage fields, on the loop you already run.\nDisclosure: I work on\nMnemoverse\n, a memory engine for AI agents connected over MCP, so weigh the argument accordingly. Every source above is linked and dated, and the longer version with the full citations is on the\noriginal page\n.",
  "metadata": {
    "title": "Rename a Tool, Lose the Entire Cache. The Vendor Documents It. - DEV Community",
    "description": "A cache discount does not reduce what a miss costs, it widens the gap: one Anthropic footnote now... Tagged with ai, agents, llm, architecture.",
    "keywords": "ai, agents, llm, architecture, software, coding, development, engineering, inclusive, community",
    "author": null,
    "og:type": "article",
    "og:url": "https://dev.to/izgorodin/cache-reads-got-cheaper-this-week-why-does-your-agent-still-pay-full-price-1p3h",
    "og:title": "Rename a Tool, Lose the Entire Cache. The Vendor Documents It.",
    "og:description": "A cache discount does not reduce what a miss costs, it widens the gap: one Anthropic footnote now...",
    "og:site_name": "DEV Community",
    "og:image": "https://media2.dev.to/dynamic/image/width=1200,height=627,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyy4awmker1hix4xsec9j.png",
    "twitter:site": "@thepracticaldev",
    "twitter:creator": "@",
    "twitter:title": "Rename a Tool, Lose the Entire Cache. The Vendor Documents It.",
    "twitter:description": "A cache discount does not reduce what a miss costs, it widens the gap: one Anthropic footnote now...",
    "twitter:card": "summary_large_image",
    "twitter:widgets:new-embed-design": "on",
    "twitter:image:src": "https://media2.dev.to/dynamic/image/width=1200,height=627,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyy4awmker1hix4xsec9j.png"
  },
  "paywall": {
    "listed_paid": false,
    "detected": false,
    "marker": "",
    "public_only": false,
    "policy": "付费/会员内容只保留页面公开可见部分，不绕过付费墙"
  },
  "crawl": {
    "success": true,
    "status_code": 200,
    "error": ""
  },
  "code_blocks": [],
  "images": [],
  "audio": [],
  "videos": [],
  "links": [
    {
      "href": "https://platform.claude.com/docs/en/about-claude/pricing",
      "label": "pricing",
      "type": "a"
    },
    {
      "href": "https://platform.openai.com/docs/pricing",
      "label": "pricing",
      "type": "a"
    },
    {
      "href": "https://ai.google.dev/gemini-api/docs/pricing",
      "label": "pricing",
      "type": "a"
    },
    {
      "href": "https://console.groq.com/docs/prompt-caching",
      "label": "caching docs",
      "type": "a"
    },
    {
      "href": "https://api-docs.deepseek.com/quick_start/pricing/",
      "label": "pricing",
      "type": "a"
    },
    {
      "href": "https://platform.claude.com/docs/en/models/fable-5-1/overview",
      "label": "model overview page",
      "type": "a"
    },
    {
      "href": "https://dev.to/max_quimby/fable-51s-real-story-is-the-75-cache-price-cut-1dh6",
      "label": "Fable 5.1's Real Story Is the 75% Cache Price Cut",
      "type": "a"
    },
    {
      "href": "https://dev.to/aidiveyt/fable-51-cut-cache-reads-by-75-cost-per-task-went-up-20-anyway-3l2i",
      "label": "Fable 5.1 cut cache reads by 75%. Cost per task went up 20% anyway.",
      "type": "a"
    },
    {
      "href": "https://platform.claude.com/docs/en/build-with-claude/prompt-caching",
      "label": "prompt caching page",
      "type": "a"
    },
    {
      "href": "https://api-docs.deepseek.com/guides/kv_cache",
      "label": "cache guide",
      "type": "a"
    },
    {
      "href": "https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus",
      "label": "engineering post of 2025-07-18",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2310.08560",
      "label": "MemGPT",
      "type": "a"
    },
    {
      "href": "https://mnemoverse.com/docs/research/agents/kv-cache-context-engineering",
      "label": "original page",
      "type": "a"
    },
    {
      "href": "https://mnemoverse.com/",
      "label": "Mnemoverse",
      "type": "a"
    }
  ],
  "repair_version": 2,
  "repaired_at": "2026-10-14T10:49:28Z",
  "source_url": "https://dev.to/izgorodin/cache-reads-got-cheaper-this-week-why-does-your-agent-still-pay-full-price-1p3h",
  "papers": [
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2310.08560",
      "bytes": 663708,
      "reused": true,
      "local_path": "papers/repair/79f95208d786cc7e6610.pdf"
    }
  ],
  "actual_video_files": [],
  "body_chars": 8454,
  "expected_body_images": 0,
  "discovery_status": "ok",
  "delivery_status": "ok",
  "missing_reason": "",
  "raw_html": "attachments/a8bcc1051de4/pages/source.html",
  "local_html": "attachments/a8bcc1051de4/pages/article.html",
  "body_base_path": "attachments/a8bcc1051de4/pages/",
  "body_markdown_path": "attachments/a8bcc1051de4/pages/article.md"
}