{
  "name": "AI Signal LLM API pricing dataset: recorded price changes",
  "publisher": "AI Signal",
  "homepage": "https://aisignalhq.com/data/",
  "generatedAt": "2026-11-06T13:21:40.152Z",
  "license": "Creative Commons Attribution 4.0 International (CC BY 4.0)",
  "licenseUrl": "https://creativecommons.org/licenses/by/4.0/",
  "attribution": "LLM API pricing data by AI Signal (https://aisignalhq.com/api-pricing/), licensed under CC BY 4.0.",
  "rowCount": 33,
  "rows": [
    {
      "id": "google-flash-promo-ends",
      "effectiveDate": "2027-01-01",
      "provider": "google",
      "kind": "provider",
      "direction": "increase",
      "multiplier": 2,
      "affects": [
        "google/gemini-3.8-flash",
        "google/gemini-3.7-flash",
        "google/gemini-3.6-flash"
      ],
      "figures": [
        {
          "model": "google/gemini-3.8-flash",
          "before": {
            "input": 0.75,
            "output": 3.75
          },
          "after": {
            "input": 1.5,
            "output": 7.5
          }
        },
        {
          "model": "google/gemini-3.7-flash",
          "before": {
            "input": 0.75,
            "output": 3.75
          },
          "after": {
            "input": 1.5,
            "output": 7.5
          }
        },
        {
          "model": "google/gemini-3.6-flash",
          "before": {
            "input": 0.75,
            "output": 3.75
          },
          "after": {
            "input": 1.5,
            "output": 7.5
          }
        }
      ],
      "summary": "Gemini 3.6, 3.7 and 3.8 Flash pricing is promotional until December 31, 2026. On January 1, 2027 it doubles: input from $0.75 to $1.50 and output from $3.75 to $7.50 per million tokens, the list prices Google shows next to the promotional ones.",
      "sourceUrl": "https://aistudio.google.com/docs/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#google-flash-promo-ends"
    },
    {
      "id": "google-gemini-4-argon-promo-ends",
      "effectiveDate": "2026-12-01",
      "provider": "google",
      "kind": "provider",
      "direction": "increase",
      "multiplier": 2,
      "affects": [
        "google/gemini-4-argon"
      ],
      "figures": [
        {
          "model": "google/gemini-4-argon",
          "before": {
            "input": 2,
            "output": 10
          },
          "after": {
            "input": 4,
            "output": 20
          }
        }
      ],
      "summary": "Gemini 4 Argon pricing is introductory until November 30, 2026. On December 1, 2026 it doubles: input from $2.00 to $4.00 and output from $10.00 to $20.00 per million tokens, the list prices Google shows next to the introductory ones.",
      "sourceUrl": "https://aistudio.google.com/docs/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#google-gemini-4-argon-promo-ends"
    },
    {
      "id": "google-gemini-omni-1-2-flash-out-of-scope",
      "effectiveDate": "2026-11-06",
      "provider": "google",
      "kind": "criterion",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "google/gemini-omni-1.2-flash"
      ],
      "figures": [],
      "summary": "Change of this site's criterion, not a price change by Google. Gemini Omni 1.2 Flash (gemini-omni-1.2-flash) was added on November 6, 2026 from Google's pricing page ($1.50 per million input tokens, $9.00 per million output tokens for text, $12.00 for video) and kept out of the table, the calculator and the counts from that day, as Gemini Omni 1.1 Flash has been since September 30, 2026: Google's model page describes 'a high-performance model for fast, conversational video generation and editing', and the $9.00 (text) figure is labelled as including thinking tokens. The page lists the model's output as 'Video with native Audio, Text'. The row stays in the dataset with its prices and dates, and its page is online, out of the search index.",
      "sourceUrl": "https://aistudio.google.com/docs/models/gemini-omni-flash",
      "pageUrl": "https://aisignalhq.com/price-changes/#google-gemini-omni-1-2-flash-out-of-scope"
    },
    {
      "id": "mistral-large-4-price-down",
      "effectiveDate": "2026-10-30",
      "provider": "mistral",
      "kind": "provider",
      "direction": "decrease",
      "multiplier": 1.9429,
      "affects": [
        "mistral/mistral-large-4"
      ],
      "figures": [
        {
          "model": "mistral/mistral-large-4",
          "before": {
            "input": 1.36,
            "output": 4.18
          },
          "after": {
            "input": 0.7,
            "output": 2.1
          }
        }
      ],
      "summary": "Mistral cut the price of Mistral Large 4, in Public Preview, from $1.36 to $0.70 per million input tokens (1.36 ÷ 0.70 = 1.94×), from $4.18 to $2.10 for output (1.99×) and from $0.14 to $0.07 for cached input. The last reading at the old price was the weekly check of October 26, 2026, 'Mistral Large 4 ↗ | $1.36 | $0.14 | $4.18 |'; on October 30, 2026 the pricing page printed 'Mistral Large 4 ↗ | $0.7 | $0.07 | $2.1 |'. Mistral does not say when it changed the price or whether the new one is temporary, so the date here is the first day it was seen.",
      "sourceUrl": "https://docs.mistral.ai/inference/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#mistral-large-4-price-down"
    },
    {
      "id": "deepinfra-cached-input-found",
      "effectiveDate": "2026-10-29",
      "provider": "deepinfra",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "deepseek-ai/DeepSeek-V4-Flash-0731",
        "moonshotai/Kimi-K3"
      ],
      "figures": [],
      "summary": "Correction of our own reading, not a price change by DeepInfra. DeepInfra's pricing page prints a cached input price inside the input cell of some models ('$0.06 / $0.015 cached'), and two offerings of this dataset did not carry it: DeepSeek V4 Flash 0731 ($0.015) and Kimi K3 ($0.285) per million tokens, read on October 29, 2026. Both rows already printed it when read on September 30, 2026, as the quotes in their notes show; the figure was not carried into the offering. Input and output prices are unchanged.",
      "sourceUrl": "https://deepinfra.com/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#deepinfra-cached-input-found"
    },
    {
      "id": "fireworks-cached-input-found",
      "effectiveDate": "2026-10-29",
      "provider": "fireworks",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "gpt-oss-120b",
        "minimax-m3"
      ],
      "figures": [],
      "summary": "Correction of our own reading, not a price change by Fireworks AI. Fireworks AI's serverless pricing page prints each Standard price as input / cached input / output, and two offerings of this dataset did not carry the cached input price: gpt-oss-120b ($0.015) and MiniMax M3 ($0.06) per million tokens, read on October 29, 2026. Both rows already printed it when read on September 30, 2026, as the quotes in their notes show; the figure was not carried into the offering. Input and output prices are unchanged.",
      "sourceUrl": "https://docs.fireworks.ai/serverless/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#fireworks-cached-input-found"
    },
    {
      "id": "openai-gpt-5-6-cyber-cache-writes-found",
      "effectiveDate": "2026-10-29",
      "provider": "openai",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "openai/gpt-5.6-cyber"
      ],
      "figures": [],
      "summary": "Correction of our own reading, not a price change by OpenAI. The Markdown version of OpenAI's pricing page (developers.openai.com/api/docs/pricing.md) prints, in the Standard table of GPT-5.6 Sol and the Cyber models, the row '| gpt-5.6-cyber | $12.50 | $75.00 | $1.25 | $15.625 | $31.25 | - | - | - | - | - |': a 30-minute cache write of $15.625 and a 12-hour cache write of $31.25 per million tokens, read on October 29, 2026. Before that this dataset showed GPT-5.6 Cyber with no cache write price; input ($12.50), cached input ($1.25) and output ($75.00) are unchanged. When OpenAI began printing these figures is not known, so the date here is the day they were read.",
      "sourceUrl": "https://developers.openai.com/api/docs/pricing.md",
      "pageUrl": "https://aisignalhq.com/price-changes/#openai-gpt-5-6-cyber-cache-writes-found"
    },
    {
      "id": "openai-gpt-5-mini-o3-price-found",
      "effectiveDate": "2026-10-26",
      "provider": "openai",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "openai/gpt-5",
        "openai/gpt-5-mini",
        "openai/o3"
      ],
      "figures": [
        {
          "model": "openai/gpt-5",
          "before": null,
          "after": {
            "input": 1.25,
            "output": 10
          }
        },
        {
          "model": "openai/gpt-5-mini",
          "before": null,
          "after": {
            "input": 0.25,
            "output": 2
          }
        },
        {
          "model": "openai/o3",
          "before": null,
          "after": {
            "input": 2,
            "output": 8
          }
        }
      ],
      "summary": "Correction of our own reading, not a price change by OpenAI. Since September 3, 2026 this site showed GPT-5, GPT-5 mini and o3 with no price, and said GPT-5's price was no longer published. OpenAI's pricing page does print them, in its folded 'All models' table, which the page draws in the browser and its Markdown version (developers.openai.com/api/docs/pricing.md) prints. Read there on October 26, 2026, the Standard table prints 'gpt-5 | $1.25 | $10.00 | $0.125', 'gpt-5-mini | $0.25 | $2.00 | $0.025' and 'o3 | $2.00 | $8.00 | $0.50' (Input, Output, Cached input, per 1M tokens); each model page, read the same day, prints the same figures. All three are deprecated and OpenAI shuts them down on December 11, 2026.",
      "sourceUrl": "https://developers.openai.com/api/docs/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#openai-gpt-5-mini-o3-price-found"
    },
    {
      "id": "openai-pricing-page-all-models-table-found",
      "effectiveDate": "2026-10-26",
      "provider": "openai",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "openai/gpt-4",
        "openai/gpt-4-turbo",
        "openai/gpt-3.5-turbo",
        "openai/o1",
        "openai/o1-pro",
        "openai/o3-mini"
      ],
      "figures": [],
      "summary": "Correction of our own reading, not a price change by OpenAI. The entry of October 5, 2026 (openai-deprecated-model-pages-price-found) said that OpenAI's pricing page no longer listed GPT-4, GPT-4 Turbo, GPT-3.5 Turbo, o1, o1-pro and o3-mini. That was this site's reading of the page, which did not see its folded 'All models' table: the page draws it in the browser, and its Markdown version (developers.openai.com/api/docs/pricing.md) prints it. Read there on October 26, 2026, three days after the six were shut down, the Standard table still lists them, at the same prices this site carries: 'gpt-4-0613 | $30.00 | $60.00 | -', 'gpt-4-turbo-2024-04-09 | $10.00 | $30.00 | -', 'gpt-3.5-turbo-0125 | $0.50 | $1.50 | -', 'o1 | $15.00 | $60.00 | $7.50', 'o1-pro | $150.00 | $600.00 | -' and 'o3-mini | $1.10 | $4.40 | $0.55' (Input, Output, Cached input, per 1M tokens). No figure shown here changes. Whether the table listed them on October 5, 2026 is not known.",
      "sourceUrl": "https://developers.openai.com/api/docs/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#openai-pricing-page-all-models-table-found"
    },
    {
      "id": "mistral-cached-input-found",
      "effectiveDate": "2026-10-22",
      "provider": "mistral",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "mistral/mistral-large-3",
        "mistral/mistral-medium-3.5",
        "mistral/mistral-small-4",
        "mistral/ministral-3-14b",
        "mistral/ministral-3-8b",
        "mistral/ministral-3-3b",
        "mistral/codestral"
      ],
      "figures": [],
      "summary": "Correction of our own reading, not a price change by Mistral. Mistral's pricing page prints a cached input price for each model in a 'Cached input' column, and seven rows of this dataset did not carry it: Mistral Large 3 ($0.05), Mistral Medium 3.5 ($0.15), Mistral Small 4 ($0.015), Ministral 3 14B ($0.02), Ministral 3 8B ($0.015), Ministral 3 3B ($0.01) and Codestral ($0.03) per million tokens, read on October 22, 2026. Before that the rows showed no cached input price; input and output prices are unchanged. When Mistral began printing these figures is not known, so the date here is the day they were read.",
      "sourceUrl": "https://docs.mistral.ai/inference/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#mistral-cached-input-found"
    },
    {
      "id": "together-kimi-k3-cached-input-up",
      "effectiveDate": "2026-10-17",
      "provider": "together",
      "kind": "provider",
      "direction": "increase",
      "multiplier": null,
      "affects": [
        "moonshotai/Kimi-K3"
      ],
      "figures": [],
      "summary": "Together AI raised the cached input price of Kimi K3 from $0.27 to $0.30 per million tokens; input ($3.00) and output ($15.00) are unchanged, so the headline price does not move. The last reading at $0.27 was AI Signal's hand reading of October 12, 2026, 'Kimi K3 | $3.00 $0.27 (cached) | $15.00 |'; AI Signal's next saved reading of the page, on October 17, 2026, already printed 'Kimi K3 | $3.00 $0.30 (cached) | $15.00 |', as did the weekly detector on October 19 and 26 and the page on October 29, 2026. The exact day of the change is not known, so the date here is the first day it was seen. Correction, October 30, 2026: this entry was dated October 19 until then; the October 17 reading had been missed.",
      "sourceUrl": "https://www.together.ai/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#together-kimi-k3-cached-input-up"
    },
    {
      "id": "meta-muse-spark-1.1-price-found",
      "effectiveDate": "2026-10-14",
      "provider": "meta",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "meta/muse-spark-1.1"
      ],
      "figures": [
        {
          "model": "meta/muse-spark-1.1",
          "before": null,
          "after": {
            "input": 1.25,
            "output": 4.25
          }
        }
      ],
      "summary": "Correction of our own reading, not a price change by Meta. Since September 16, 2026 this site showed Muse Spark 1.1 as 'no published price', because its own model page carries a benchmark table and no pricing table. On October 14, 2026 Meta's 'Pricing and rate limits' page, which this site had not read, lists muse-spark-1.1 in Standard mode next to 1.3 and 1.2 ('Muse Spark 1.1 is Standard mode only.') at $1.25 input, $0.15 cached input and $4.25 output per million tokens. The row now carries that price. When Meta began listing 1.1 on that page is not known, so the date here is the day it was read.",
      "sourceUrl": "https://dev.meta.ai/docs/pricing-rate-limits",
      "pageUrl": "https://aisignalhq.com/price-changes/#meta-muse-spark-1.1-price-found"
    },
    {
      "id": "google-gemini-3-flash-preview-output-added",
      "effectiveDate": "2026-10-13",
      "provider": "google",
      "kind": "omission",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "google/gemini-3-flash-preview"
      ],
      "figures": [],
      "summary": "Omission on our side, not a price change by Google. The Gemini 3 Flash Preview row had an input price ($0.50) and a cache price ($0.05) per million tokens but no output price, because it was read on Google Cloud's Vertex AI pricing page, whose standard table prints only input rows for this model. On October 13, 2026 Google's Gemini API pricing page, the source of every other Gemini row, was read and prints an output price of $3.00 per million tokens (including thinking tokens) for the same model. The row now cites that page, with $0.50 input, $3.00 output and $0.05 cached. This site does not know since when the Gemini API page has printed that figure.",
      "sourceUrl": "https://aistudio.google.com/docs/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#google-gemini-3-flash-preview-output-added"
    },
    {
      "id": "together-kimi-k3-promo-read-as-list",
      "effectiveDate": "2026-10-12",
      "provider": "together",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "moonshotai/Kimi-K3"
      ],
      "figures": [
        {
          "model": "moonshotai/Kimi-K3",
          "before": {
            "input": 2.7,
            "output": 13.5
          },
          "after": {
            "input": 3,
            "output": 15
          }
        }
      ],
      "summary": "Correction of our own reading, not a price change by Together AI. From October 6 to October 12, 2026 this site published Kimi K3 on Together AI at $2.70 input and $13.50 output per million tokens and recorded it as a 10% price cut by Together. It was a promotion: on October 12, 2026 Together's pricing page labels the row 'PROMO' with 'Promotion valid until October 11, 2026' and prints the list price beside it, $3.00 input and $15.00 output, the price it had until October 5. The list price did not change and this site publishes list prices, so the row is back at $3.00 / $15.00. Cached input stays at $0.27, the figure Together's Standard table prints on October 12, 2026 (it was $0.30 until October 5). This entry replaces the one of October 6 that called the promotion a price cut.",
      "sourceUrl": "https://www.together.ai/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#together-kimi-k3-promo-read-as-list"
    },
    {
      "id": "google-gemini-3-7-flash-promo-unverifiable",
      "effectiveDate": "2026-10-11",
      "provider": "google",
      "kind": "criterion",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "google/gemini-3.7-flash"
      ],
      "figures": [],
      "summary": "Change of this site's labelling, not a price change by Google. Google deprecated Gemini 3.7 Flash, routes its requests to Gemini 3.8 Flash and no longer lists it on its pricing page, so the promotion this site showed for it ($0.75 input and $3.75 output through December 31, 2026, then $1.50 and $7.50) can no longer be checked. From October 11, 2026 the row shows the last prices read, on October 5, 2026, without the promotional label or the list price after it. The announced January 1, 2027 increase still stands for Gemini 3.6 and 3.8 Flash, which Google still prices.",
      "sourceUrl": "https://aistudio.google.com/docs/models/gemini-3.7-flash",
      "pageUrl": "https://aisignalhq.com/price-changes/#google-gemini-3-7-flash-promo-unverifiable"
    },
    {
      "id": "anthropic-sonnet-5-5-cache-read-down",
      "effectiveDate": "2026-10-07",
      "provider": "anthropic",
      "kind": "provider",
      "direction": "decrease",
      "multiplier": null,
      "affects": [
        "anthropic/claude-sonnet-5-5"
      ],
      "figures": [],
      "summary": "Anthropic halved the price of prompt cache reads on Claude Sonnet 5.5, from $0.20 to $0.10 per million tokens (0.05x the base input price instead of 0.1x), announced in its platform release notes on October 7, 2026. Input ($2), output ($10) and cache writes ($2.50 for 5 minutes, $4 for 1 hour) are unchanged, so the headline price does not move. This site read the new figure on October 11, 2026.",
      "sourceUrl": "https://platform.claude.com/docs/en/release-notes/overview",
      "pageUrl": "https://aisignalhq.com/price-changes/#anthropic-sonnet-5-5-cache-read-down"
    },
    {
      "id": "openai-deprecated-model-pages-price-found",
      "effectiveDate": "2026-10-05",
      "provider": "openai",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "openai/gpt-4",
        "openai/gpt-4-turbo",
        "openai/gpt-3.5-turbo",
        "openai/o1",
        "openai/o1-pro",
        "openai/o3-mini"
      ],
      "figures": [
        {
          "model": "openai/gpt-4",
          "before": null,
          "after": {
            "input": 30,
            "output": 60
          }
        },
        {
          "model": "openai/gpt-4-turbo",
          "before": null,
          "after": {
            "input": 10,
            "output": 30
          }
        },
        {
          "model": "openai/gpt-3.5-turbo",
          "before": null,
          "after": {
            "input": 0.5,
            "output": 1.5
          }
        },
        {
          "model": "openai/o1",
          "before": null,
          "after": {
            "input": 15,
            "output": 60
          }
        },
        {
          "model": "openai/o1-pro",
          "before": null,
          "after": {
            "input": 150,
            "output": 600
          }
        },
        {
          "model": "openai/o3-mini",
          "before": null,
          "after": {
            "input": 1.1,
            "output": 4.4
          }
        }
      ],
      "summary": "Correction of our own reading, not a price change by OpenAI. Until October 5, 2026 this site showed GPT-4, GPT-4 Turbo, GPT-3.5 Turbo, o1, o1-pro and o3-mini with no price, because it read OpenAI's pricing page, which no longer lists them, and not their model pages, which still do. Read by hand on October 5, 2026, with the page's 'Batch API price' switch off (Standard prices), the model pages print, per 1M tokens: GPT-4 'Input $30.00 Output $60.00'; GPT-4 Turbo 'Input $10.00 Output $30.00'; GPT-3.5 Turbo 'Input $0.50 Output $1.50'; o1 'Input $15.00 Cached input $7.50 Output $60.00'; o1-pro 'Input $150.00 Output $600.00'; o3-mini 'Input $1.10 Cached input $0.55 Output $4.40'. All six are deprecated and OpenAI shuts them down on October 23, 2026; each row now cites its model page.",
      "sourceUrl": "https://developers.openai.com/api/docs/deprecations",
      "pageUrl": "https://aisignalhq.com/price-changes/#openai-deprecated-model-pages-price-found"
    },
    {
      "id": "anthropic-haiku-4-5-release-date-correction",
      "effectiveDate": "2026-10-01",
      "provider": "anthropic",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "anthropic/claude-haiku-4-5"
      ],
      "figures": [],
      "summary": "Correction of this site's record, not a price change by Anthropic: no price moved. This page gave Claude Haiku 4.5's release date as October 1, 2025; Anthropic's model page, read on October 1, 2026, prints 'Released October 15, 2025', which also matches the retirement commitment on the same page ('Not sooner than October 15, 2026'). The earlier date had no source recorded. The release date shown here was corrected that day.",
      "sourceUrl": "https://platform.claude.com/docs/en/models/haiku-4-5/overview",
      "pageUrl": "https://aisignalhq.com/price-changes/#anthropic-haiku-4-5-release-date-correction"
    },
    {
      "id": "anthropic-opus-4-5-release-date-correction",
      "effectiveDate": "2026-09-30",
      "provider": "anthropic",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "anthropic/claude-opus-4-5"
      ],
      "figures": [],
      "summary": "Correction of this site's record, not a price change by Anthropic: no price moved. This page gave Claude Opus 4.5's release date as November 1, 2025; Anthropic's model page, read on September 30, 2026, prints 'Released November 24, 2025', which also matches the retirement commitment on the same page ('Not sooner than November 24, 2026'). The earlier date had no source recorded. The release date shown here was corrected that day.",
      "sourceUrl": "https://platform.claude.com/docs/en/models/opus-4-5/overview",
      "pageUrl": "https://aisignalhq.com/price-changes/#anthropic-opus-4-5-release-date-correction"
    },
    {
      "id": "deepinfra-deepseek-v4-flash-0731-input-down",
      "effectiveDate": "2026-09-30",
      "provider": "deepinfra",
      "kind": "provider",
      "direction": "decrease",
      "multiplier": 1.3333,
      "affects": [
        "deepseek-ai/DeepSeek-V4-Flash-0731"
      ],
      "figures": [
        {
          "model": "deepseek-ai/DeepSeek-V4-Flash-0731",
          "before": {
            "input": 0.08,
            "output": 0.18
          },
          "after": {
            "input": 0.06,
            "output": 0.18
          }
        }
      ],
      "summary": "DeepInfra lowered the input price of DeepSeek V4 Flash 0731 from $0.08 to $0.06 per million tokens; output stays at $0.18. It was $0.08 when this site read DeepInfra's pricing page on September 6, 2026 and $0.06 when it read it again on September 30, 2026; the exact day of the change is not known, so the date here is the day it was seen.",
      "sourceUrl": "https://deepinfra.com/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#deepinfra-deepseek-v4-flash-0731-input-down"
    },
    {
      "id": "google-gemini-omni-flash-out-of-scope",
      "effectiveDate": "2026-09-30",
      "provider": "google",
      "kind": "criterion",
      "direction": "withdrawn",
      "multiplier": null,
      "affects": [
        "google/gemini-omni-1.1-flash",
        "google/gemini-omni-flash-preview"
      ],
      "figures": [
        {
          "model": "google/gemini-omni-1.1-flash",
          "before": {
            "input": 1.5,
            "output": 9
          },
          "after": null
        },
        {
          "model": "google/gemini-omni-flash-preview",
          "before": {
            "input": 1.5,
            "output": 9
          },
          "after": null
        }
      ],
      "summary": "Change of this site's criterion, not a price change by Google. Gemini Omni Flash (gemini-omni-1.1-flash) and its preview endpoint (gemini-omni-flash-preview) entered this dataset in the hand sweep of September 16, 2026, priced $1.50 per million input tokens and $9.00 per million output tokens. On September 30, 2026 Google's model page for both versions was read: it describes a model 'for fast, conversational video generation and editing' and gives its supported output as 'Video' only. The pricing table covers models charged per token with text output, so both rows left the table, the calculator and the counts that day. The $9.00 (text) figure on Google's pricing page is labelled as including thinking tokens; it is not the price of a text answer comparable with the other rows. Both rows stay in the dataset with their prices and dates, and their pages remain online, out of the search index.",
      "sourceUrl": "https://aistudio.google.com/docs/models/gemini-omni-flash",
      "pageUrl": "https://aisignalhq.com/price-changes/#google-gemini-omni-flash-out-of-scope"
    },
    {
      "id": "alibaba-qwen3.7-flash-added",
      "effectiveDate": "2026-09-17",
      "provider": "alibaba",
      "kind": "omission",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "alibaba/qwen3.7-flash"
      ],
      "figures": [
        {
          "model": "alibaba/qwen3.7-flash",
          "before": null,
          "after": {
            "input": 0.03,
            "output": 0.13
          }
        }
      ],
      "summary": "Omission on our side, not a price change by Alibaba. Qwen3.7 Flash belongs to the same generation as the Qwen3.7 Plus and Max rows and had no row in this dataset. It turned up on September 17, 2026 through the OpenRouter signal of this site's read-only scan for missing models, and its prices were read that day on Alibaba's own Model Studio page: $0.03 input, $0.13 output and $0.006 cached per million tokens in the Singapore (International) region for prompts up to 32K tokens, $0.10 / $0.40 above 32K and $0.20 / $0.80 above 256K. For Alibaba the table covers the current Qwen generations, 3.7 and 3.8; the older generations Alibaba still sells, found the same day, were left out on purpose.",
      "sourceUrl": "https://www.alibabacloud.com/help/en/model-studio/qwen3-7-flash",
      "pageUrl": "https://aisignalhq.com/price-changes/#alibaba-qwen3.7-flash-added"
    },
    {
      "id": "meta-muse-spark-price-found",
      "effectiveDate": "2026-09-16",
      "provider": "meta",
      "kind": "correction",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "meta/muse-spark-1.3",
        "meta/muse-spark-1.2"
      ],
      "figures": [
        {
          "model": "meta/muse-spark-1.3",
          "before": null,
          "after": {
            "input": 1.25,
            "output": 4.25
          }
        },
        {
          "model": "meta/muse-spark-1.2",
          "before": null,
          "after": {
            "input": 1.25,
            "output": 4.25
          }
        }
      ],
      "summary": "Correction of our own reading, not a price change by Meta. From September 11 to September 16, 2026 this site showed Muse Spark 1.3 and Muse Spark 1.2 as 'no published price', because Meta's model pages serve their 'Models and pricing' table only to a browser, and the HTTP client this site verifies with received a JavaScript shell with no pricing text. Read by hand in a browser on September 16, 2026, each page lists two SKUs per model: 'Not used to improve our products' at $1.25 input, $0.15 cached input and $4.25 output per million tokens, and a 'contributor' SKU, 'Used to improve our products', at $0.10 / $0.002 / $0.20, conditional on letting Meta use the traffic to improve its products. The rows now carry the first as the list price, with the contributor price quoted in their notes; Meta changed nothing. Muse Spark 1.1's own model page, read the same day in a browser, carries a benchmark table and no pricing table, and neither of the other two tables lists it, so that row still shows no published price.",
      "sourceUrl": "https://developer.meta.com/ai/models/muse-spark/",
      "pageUrl": "https://aisignalhq.com/price-changes/#meta-muse-spark-price-found"
    },
    {
      "id": "sweep-2026-09-16-missing-models",
      "effectiveDate": "2026-09-16",
      "provider": "multiple",
      "kind": "omission",
      "direction": "none",
      "multiplier": null,
      "affects": [
        "openai/gpt-6-astra",
        "openai/gpt-rosalind-research",
        "anthropic/claude-haiku-3-5",
        "google/gemini-2.5-computer-use-preview",
        "google/gemini-robotics-er-2-preview",
        "google/gemini-robotics-er-2-streaming-preview",
        "google/gemini-omni-flash-preview",
        "mistral/glm-5.3"
      ],
      "figures": [],
      "summary": "Omission on our side, not a price change by any provider. On September 16, 2026, while reading OpenAI's pricing page by hand for a check on batch and flex prices, GPT-6 Astra turned up in the Standard table with no row in this dataset. A hand sweep of the first-party pricing pages of OpenAI, Anthropic, Google, xAI, Mistral, DeepSeek, Alibaba and Cohere against the dataset then found eight models charged per token with text output and a published price that had no row: GPT-6 Astra and GPT-Rosalind (OpenAI), Claude Haiku 3.5 (Anthropic, retired on the first-party API but still priced), Gemini 2.5 Computer Use Preview, Gemini Robotics ER 2 Preview and its Streaming variant, and Gemini Omni Flash Preview, a separate endpoint id from the generally available Omni Flash row at the same prices (Google), and GLM 5.3 (served by Mistral). Astra had been in OpenAI's Standard table at least since September 10, 2026, when the weekly detector first received that page's text, and this site's OpenRouter explainer already named it from the September 15 snapshot. The weekly detector could not have caught any of these: it compares the rows that exist against the page and has no step for models the page lists and the dataset does not. Each new row carries the page quote it was read from. Audio, image, video, embedding and per-minute models found in the same sweep were left out on purpose: the table covers models charged per token with text output.",
      "sourceUrl": "https://developers.openai.com/api/docs/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#sweep-2026-09-16-missing-models"
    },
    {
      "id": "openrouter-deepseek-v4-pro-tariff-rule",
      "effectiveDate": "2026-09-14",
      "provider": "openrouter",
      "kind": "criterion",
      "direction": "increase",
      "multiplier": 2,
      "affects": [
        "deepseek/deepseek-v4-pro-0813"
      ],
      "figures": [
        {
          "model": "deepseek/deepseek-v4-pro-0813",
          "before": {
            "input": 0.66,
            "output": 1.98
          },
          "after": {
            "input": 1.32,
            "output": 3.96
          }
        }
      ],
      "summary": "The figure this site publishes for DeepSeek V4 Pro 0813 on OpenRouter changes from $0.66 / $1.98 to $1.32 / $3.96 per million tokens. This is a correction of our own reading, not a price change by DeepSeek or by any host. DeepSeek's own pricing page defines $1.32 / $3.96 as the peak rate and $0.66 / $1.98 as the off-peak rate, half of it, applying outside Monday to Friday 01:00-04:00 and 06:00-10:00 UTC. Every reading behind the old figure had been taken off-peak. Read-only probes on September 11 and 14, 2026 (02:30, 07:30, 12:30 UTC) showed that DeepSeek's endpoint on OpenRouter follows that clock: $1.32 / $3.96 during peak hours, $0.66 / $1.98 the rest of the time. Since this site publishes the maker's list price without discounts, the row now shows the peak rate, which is also what Together AI and Fireworks AI charge at all hours; the off-peak rate is recorded on the row and shown as a note. The peak rate applies 35 of the 168 hours in a week.",
      "sourceUrl": "https://api-docs.deepseek.com/quick_start/pricing/",
      "pageUrl": "https://aisignalhq.com/price-changes/#openrouter-deepseek-v4-pro-tariff-rule"
    },
    {
      "id": "alibaba-reference-region-singapore",
      "effectiveDate": "2026-09-11",
      "provider": "alibaba",
      "kind": "criterion",
      "direction": "increase",
      "multiplier": null,
      "affects": [
        "alibaba/qwen3.8-max",
        "alibaba/qwen3.8-flash",
        "alibaba/qwen3.7-plus",
        "alibaba/qwen3.7-max"
      ],
      "figures": [
        {
          "model": "alibaba/qwen3.8-max",
          "before": {
            "input": 1.65,
            "output": 4.951
          },
          "after": {
            "input": 2,
            "output": 6
          }
        },
        {
          "model": "alibaba/qwen3.8-flash",
          "before": {
            "input": 0.113,
            "output": 0.382
          },
          "after": {
            "input": 0.15,
            "output": 0.47
          }
        },
        {
          "model": "alibaba/qwen3.7-plus",
          "before": {
            "input": 0.276,
            "output": 1.101
          },
          "after": {
            "input": 0.4,
            "output": 1.6
          }
        },
        {
          "model": "alibaba/qwen3.7-max",
          "before": {
            "input": 1.65,
            "output": 4.951
          },
          "after": {
            "input": 2.5,
            "output": 7.5
          }
        }
      ],
      "summary": "Change of criterion on our side, not a price change by Alibaba: no Alibaba price rose. Alibaba Cloud's Model Studio pages list the same model at two price levels: the China (Beijing) region and the Global regions (Frankfurt, Virginia, Tokyo, Hong Kong) share one level, and the Singapore region, labelled by Alibaba \"Scope: International\", is higher. Until September 11, 2026 the first-party table quoted the Beijing/Global level while the provider comparison quoted Alibaba's OpenRouter endpoint at the Singapore level, so the site showed two different \"Alibaba\" prices for the same model. From September 11 the reference region is Singapore (International), the one Alibaba labels international, the one every reseller we track (OpenRouter, Together, Fireworks) mirrors, and the higher of the two; the Beijing/Global figures stay published next to it. Published figures move from 1.65 / 4.951 to 2.00 / 6.00 (Qwen3.8 Max), 0.113 / 0.382 to 0.15 / 0.47 (Qwen3.8 Flash), 0.276 / 1.101 to 0.40 / 1.60 (Qwen3.7 Plus, prompts up to 256K; the tier above 256K, 1.20 / 4.80, is now recorded too) and 1.65 / 4.951 to 2.50 / 7.50 (Qwen3.7 Max). Alibaba's pages state that they show the original pricing excluding any limited-time promotions.",
      "sourceUrl": "https://www.alibabacloud.com/help/en/model-studio/qwen3-8-max",
      "pageUrl": "https://aisignalhq.com/price-changes/#alibaba-reference-region-singapore"
    },
    {
      "id": "openrouter-deepseek-v4-flash-maker-endpoint-withdrawn",
      "effectiveDate": "2026-09-11",
      "provider": "openrouter",
      "kind": "provider",
      "direction": "withdrawn",
      "multiplier": null,
      "affects": [
        "deepseek/deepseek-v4-flash-0731"
      ],
      "figures": [
        {
          "model": "deepseek/deepseek-v4-flash-0731",
          "before": {
            "input": 0.22,
            "output": 0.66
          },
          "after": null
        }
      ],
      "summary": "DeepSeek's own endpoint for DeepSeek V4 Flash 0731 disappeared from OpenRouter between the afternoon of September 9, 2026, when it was still listed at $0.22 / $0.66 and this row was verified against it, and the 04:30 UTC check of September 10, which found no DeepSeek endpoint; it had not returned by September 11. DeepSeek still serves V4 Pro 0813 on OpenRouter, so this is specific to this model, not a general withdrawal. This site's OpenRouter column shows the model maker's list price, and for this model there is now none to show: the column moves from $0.22 / $0.66 per million tokens (last verified September 9) to \"no list price\". Nothing changed in our criterion, and no host changed a price; the provider withdrew its endpoint. The model is still served on OpenRouter by resellers (28 listed, 26 healthy on September 11), reachable with :floor; that day the catalog quoted $0.065 / $0.18 via Relace.",
      "sourceUrl": "https://openrouter.ai/api/v1/models/deepseek/deepseek-v4-flash-0731/endpoints",
      "pageUrl": "https://aisignalhq.com/price-changes/#openrouter-deepseek-v4-flash-maker-endpoint-withdrawn"
    },
    {
      "id": "openai-gpt56-promo-ends",
      "effectiveDate": "2026-09-10",
      "provider": "openai",
      "kind": "correction",
      "direction": "increase",
      "multiplier": 2,
      "affects": [
        "openai/gpt-5.6-sol",
        "openai/gpt-5.6-terra",
        "openai/gpt-5.6-luna"
      ],
      "figures": [
        {
          "model": "openai/gpt-5.6-sol",
          "before": {
            "input": 2,
            "output": 10
          },
          "after": {
            "input": 4,
            "output": 20
          }
        },
        {
          "model": "openai/gpt-5.6-terra",
          "before": {
            "input": 1,
            "output": 6
          },
          "after": {
            "input": 2,
            "output": 12
          }
        },
        {
          "model": "openai/gpt-5.6-luna",
          "before": {
            "input": 0.1,
            "output": 0.6
          },
          "after": {
            "input": 0.2,
            "output": 1.2
          }
        }
      ],
      "summary": "Correction of our own reading, not a price change by OpenAI. From September 3 to September 10, 2026 this site published GPT-5.6 Sol at $2 / $10, Terra at $1 / $6 and Luna at $0.10 / $0.60 per million tokens, read from the Batch tab of OpenAI's pricing page, which is always 50% of Standard (the page's four tabs, Standard, Batch, Flex and Fast mode, look identical in plain text), and labelled as a promotion ending November 21 with double the price after it. The Standard short-context prices, verified by hand on September 10, are Sol $4 / $20, Terra $2 / $12 and Luna $0.20 / $1.20, and the published figures doubled to them. The November 21 date was OpenAI's, but it belongs to Sol alone and is not an end date: OpenAI states that Sol's $4 / $20 is promotional pricing, a 20% reduction on input and 33% on output, available at least through November 21, 2026, and publishes no list price for after it; Terra and Luna carry no promotion. On September 10 this entry wrongly said there was no promotion to end; corrected on September 15, when the Sol row got its promotion back, marked as at least through that date, without an invented list price.",
      "sourceUrl": "https://developers.openai.com/api/docs/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#openai-gpt56-promo-ends"
    },
    {
      "id": "openrouter-deepseek-v4-flash-list-price-rule",
      "effectiveDate": "2026-09-09",
      "provider": "openrouter",
      "kind": "criterion",
      "direction": "increase",
      "multiplier": 4.4,
      "affects": [
        "deepseek/deepseek-v4-flash-0731"
      ],
      "figures": [
        {
          "model": "deepseek/deepseek-v4-flash-0731",
          "before": {
            "input": 0.05,
            "output": 0.16
          },
          "after": {
            "input": 0.22,
            "output": 0.66
          }
        }
      ],
      "summary": "The published OpenRouter price of DeepSeek V4 Flash 0731 rises 4.4x on input ($0.05 to $0.22) and 4.1x on output ($0.16 to $0.66). No host raised its rate. The number changes because the rule changed: until September 9, 2026 this column showed the cheapest healthy endpoint at the moment of checking (OpenInference); from September 9 it shows the model maker's own price on OpenRouter's providers page, without promotional discounts (DeepSeek, $0.22 / $0.66). Cheaper resellers still exist and are still reachable with :floor.",
      "sourceUrl": "https://openrouter.ai/api/v1/models/deepseek/deepseek-v4-flash-0731/endpoints",
      "pageUrl": "https://aisignalhq.com/price-changes/#openrouter-deepseek-v4-flash-list-price-rule"
    },
    {
      "id": "openrouter-glm-5.3-flash-list-price-rule",
      "effectiveDate": "2026-09-09",
      "provider": "openrouter",
      "kind": "criterion",
      "direction": "increase",
      "multiplier": 2.1,
      "affects": [
        "z-ai/glm-5.3-flash"
      ],
      "figures": [
        {
          "model": "z-ai/glm-5.3-flash",
          "before": {
            "input": 0.0713,
            "output": 0.2375
          },
          "after": {
            "input": 0.15,
            "output": 0.5
          }
        }
      ],
      "summary": "The published OpenRouter price of GLM-5.3 Flash rises 2.1x on both input ($0.0713 to $0.15) and output ($0.2375 to $0.50). No host raised its rate. The number changes because the rule changed: until September 9, 2026 this column showed the cheapest healthy endpoint at the moment of checking (Relace); from September 9 it shows the model maker's list price on OpenRouter's providers page, without promotional discounts (Z.ai, $0.15 / $0.50). On September 9 Z.ai was still showing a 50% launch discount ($0.075 / $0.25, struck-through list price) that it had announced would end at midnight Singapore time, 16:00 UTC. Cheaper resellers still exist and are still reachable with :floor.",
      "sourceUrl": "https://openrouter.ai/api/v1/models/z-ai/glm-5.3-flash/endpoints",
      "pageUrl": "https://aisignalhq.com/price-changes/#openrouter-glm-5.3-flash-list-price-rule"
    },
    {
      "id": "openrouter-glm-5.3-list-price-rule",
      "effectiveDate": "2026-09-09",
      "provider": "openrouter",
      "kind": "criterion",
      "direction": "increase",
      "multiplier": 1.25,
      "affects": [
        "z-ai/glm-5.3"
      ],
      "figures": [
        {
          "model": "z-ai/glm-5.3",
          "before": {
            "input": 1.12,
            "output": 3.52
          },
          "after": {
            "input": 1.4,
            "output": 4.4
          }
        }
      ],
      "summary": "The published OpenRouter price of GLM-5.3 rises 1.25x on both input ($1.12 to $1.40) and output ($3.52 to $4.40). No host raised its rate. The number changes because the rule changed: until September 9, 2026 this column showed the cheapest healthy endpoint at the moment of checking (GMICloud); from September 9 it shows the model maker's own price on OpenRouter's providers page, without promotional discounts (Z.ai, $1.40 / $4.40). Cheaper resellers still exist and are still reachable with :floor.",
      "sourceUrl": "https://openrouter.ai/api/v1/models/z-ai/glm-5.3/endpoints",
      "pageUrl": "https://aisignalhq.com/price-changes/#openrouter-glm-5.3-list-price-rule"
    },
    {
      "id": "openrouter-deepseek-v4-pro-cheapest-host-changes",
      "effectiveDate": "2026-09-08",
      "provider": "openrouter",
      "kind": "criterion",
      "direction": "increase",
      "multiplier": 1.136,
      "affects": [
        "deepseek/deepseek-v4-pro-0813"
      ],
      "figures": [
        {
          "model": "deepseek/deepseek-v4-pro-0813",
          "before": {
            "input": 0.5808,
            "output": 1.7424
          },
          "after": {
            "input": 0.66,
            "output": 1.98
          }
        }
      ],
      "summary": "The figure this site published for DeepSeek V4 Pro 0813 on OpenRouter rose 13.6% on September 8, 2026, from $0.5808 / $1.7424 to $0.66 / $1.98 per million tokens. DeepSeek did not change any price. At the time the published figure was the cheapest healthy OpenRouter endpoint, and that endpoint switched from Alibaba to DeepSeek's own: DeepSeek's endpoint read $0.66 / $1.98 on both days, while Alibaba's went from $0.5808 / $1.7424 to $1.122 / $3.366 and has since been seen at both figures again. This entry originally described Alibaba's move as a 12% discount ending; that reading was wrong. Both readings were also taken outside DeepSeek's peak hours, so $0.66 / $1.98 was the maker's off-peak rate, not its list price. This rule was replaced on September 9 (maker's list price) and again on September 14 (see openrouter-deepseek-v4-pro-tariff-rule). Kept because the number on this site did change that day; it was our rule, not a host's price.",
      "sourceUrl": "https://openrouter.ai/api/v1/models/deepseek/deepseek-v4-pro-0813/endpoints",
      "pageUrl": "https://aisignalhq.com/price-changes/#openrouter-deepseek-v4-pro-cheapest-host-changes"
    },
    {
      "id": "fireworks-us-only-serverless-surcharge",
      "effectiveDate": "2026-09-01",
      "provider": "fireworks",
      "kind": "provider",
      "direction": "increase",
      "multiplier": 1.5,
      "affects": [],
      "figures": [],
      "summary": "Since September 1, 2026 Fireworks adds a 50% surcharge to the base price of the Serverless models it marks as US-only. The Fireworks prices on this site are the base prices, without the surcharge.",
      "sourceUrl": "https://docs.fireworks.ai/serverless/pricing",
      "pageUrl": "https://aisignalhq.com/price-changes/#fireworks-us-only-serverless-surcharge"
    }
  ]
}