[{"data":1,"prerenderedAt":1907},["ShallowReactive",2],{"navigation_docs":3,"-observability-metrics":160,"-observability-metrics-surround":1902},[4,21,62,82,103,149],{"title":5,"icon":6,"path":7,"stem":8,"children":9,"page":20},"Get Started","i-lucide-rocket","\u002Fgetting-started","1.getting-started",[10,15],{"title":11,"path":12,"stem":13,"icon":14},"Introduction","\u002Fgetting-started\u002Fintroduction","1.getting-started\u002F1.introduction","i-lucide-info",{"title":16,"path":17,"stem":18,"icon":19},"Quickstart","\u002Fgetting-started\u002Fquickstart","1.getting-started\u002F2.quickstart","i-lucide-play-circle",false,{"title":22,"icon":23,"path":24,"stem":25,"children":26,"page":20},"Gateway","i-lucide-network","\u002Fgateway","2.gateway",[27,32,37,42,47,52,57],{"title":28,"path":29,"stem":30,"icon":31},"Overview","\u002Fgateway\u002Foverview","2.gateway\u002F1.overview","i-lucide-waypoints",{"title":33,"path":34,"stem":35,"icon":36},"Models and Routing","\u002Fgateway\u002Fmodels-and-routing","2.gateway\u002F2.models-and-routing","i-lucide-brain",{"title":38,"path":39,"stem":40,"icon":41},"Smart Router","\u002Fgateway\u002Fsmart-routing","2.gateway\u002F3.smart-routing","i-lucide-route",{"title":43,"path":44,"stem":45,"icon":46},"Context Engine","\u002Fgateway\u002Fcontext-engine","2.gateway\u002F4.context-engine","i-product-context-engine",{"title":48,"path":49,"stem":50,"icon":51},"Data Integrations","\u002Fgateway\u002Fdata-integrations","2.gateway\u002F5.data-integrations","i-lucide-database",{"title":53,"path":54,"stem":55,"icon":56},"Tools","\u002Fgateway\u002Ftools","2.gateway\u002F6.tools","i-lucide-wrench",{"title":58,"path":59,"stem":60,"icon":61},"Guardrails","\u002Fgateway\u002Fguardrails","2.gateway\u002F7.guardrails","i-lucide-shield-alert",{"title":63,"icon":64,"path":65,"stem":66,"children":67,"page":20},"Compliance","i-lucide-shield-check","\u002Fcompliance","3.compliance",[68,72,77],{"title":69,"path":70,"stem":71,"icon":64},"Data Governance","\u002Fcompliance\u002Fdata-governance","3.compliance\u002F01.data-governance",{"title":73,"path":74,"stem":75,"icon":76},"Network Access","\u002Fcompliance\u002Fnetwork-access","3.compliance\u002F02.network-access","i-lucide-lock",{"title":78,"path":79,"stem":80,"icon":81},"Audit Logs","\u002Fcompliance\u002Faudit-logs","3.compliance\u002F03.audit-logs","i-lucide-scroll-text",{"title":83,"icon":84,"path":85,"stem":86,"children":87,"page":20},"Observability","i-lucide-eye","\u002Fobservability","4.observability",[88,93,98],{"title":89,"path":90,"stem":91,"icon":92},"Organization","\u002Fobservability\u002Forganization","4.observability\u002F1.organization","i-lucide-building-2",{"title":94,"path":95,"stem":96,"icon":97},"Application","\u002Fobservability\u002Fapplication","4.observability\u002F2.application","i-lucide-activity",{"title":99,"path":100,"stem":101,"icon":102},"Metrics","\u002Fobservability\u002Fmetrics","4.observability\u002F3.metrics","i-lucide-bar-chart-2",{"title":104,"icon":105,"path":106,"stem":107,"children":108,"page":20},"Administration","i-lucide-settings","\u002Fadministration","5.administration",[109,114,119,129,144],{"title":110,"path":111,"stem":112,"icon":113},"Organization and Projects","\u002Fadministration\u002Forganizations-and-projects","5.administration\u002F1.organizations-and-projects","i-lucide-box",{"title":115,"path":116,"stem":117,"icon":118},"Members and Access","\u002Fadministration\u002Fmembers-and-access","5.administration\u002F2.members-and-access","i-lucide-shield-user",{"title":120,"path":121,"stem":122,"children":123,"icon":76},"Authentication","\u002Fadministration\u002Fauthentication","5.administration\u002F3.authentication",[124],{"title":125,"path":126,"stem":127,"icon":128},"Connect Model APIs","\u002Fadministration\u002Fauthentication\u002Fconnect-model-apis","5.administration\u002F3.authentication\u002F31.connect-model-apis","i-lucide-cpu",{"title":130,"path":131,"stem":132,"children":133,"icon":143},"Connections","\u002Fadministration\u002Fconnections","5.administration\u002F4.connections",[134,139],{"title":135,"path":136,"stem":137,"icon":138},"Connect MCP Servers","\u002Fadministration\u002Fconnections\u002Fmcp-connections","5.administration\u002F4.connections\u002F42.mcp-connections","i-lucide-plug",{"title":140,"path":141,"stem":142,"icon":51},"Connect Data Sources","\u002Fadministration\u002Fconnections\u002Fdata-sources-connections","5.administration\u002F4.connections\u002F43.data-sources-connections","i-lucide-link",{"title":145,"path":146,"stem":147,"icon":148},"Budgets","\u002Fadministration\u002Fbudgets","5.administration\u002F5.budgets","i-lucide-wallet",{"title":150,"icon":151,"path":152,"stem":153,"children":154,"page":20},"Reference","i-lucide-book-open","\u002Freference","6.reference",[155],{"title":156,"path":157,"stem":158,"icon":159},"Using the API","\u002Freference\u002Fapi-usage","6.reference\u002F1.api-usage","i-lucide-terminal-square",{"id":161,"title":99,"body":162,"description":1895,"extension":1896,"links":1897,"meta":1898,"navigation":1899,"path":100,"seo":1900,"stem":101,"__hash__":1901},"docs\u002F4.observability\u002F3.metrics.md",{"type":163,"value":164,"toc":1875},"minimark",[165,177,182,190,281,286,289,327,331,342,359,362,383,386,391,396,400,535,539,636,643,647,650,672,678,682,689,769,772,1001,1004,1007,1026,1029,1232,1235,1239,1249,1427,1431,1437,1565,1568,1574,1664,1667,1682,1841,1845,1859,1863],[166,167,168,169,173,174,176],"p",{},"This page defines every number on the ",[170,171,172],"strong",{},"Dashboard"," and on an application's ",[170,175,28],{},", so you know what each one counts and how Optiak calculates it. The metric names match the labels in the console.",[178,179,181],"h2",{"id":180},"how-metrics-are-calculated","How metrics are calculated",[166,183,184,185,173,187,189],{},"These rules apply to every metric on the ",[170,186,172],{},[170,188,28],{},".",[191,192,193,206],"table",{},[194,195,196],"thead",{},[197,198,199,203],"tr",{},[200,201,202],"th",{},"Rule",[200,204,205],{},"What it means",[207,208,209,235,251,261,271],"tbody",{},[197,210,211,217],{},[212,213,214],"td",{},[170,215,216],{},"Period",[212,218,219,220,223,224,223,227,230,231,234],{},"You select ",[170,221,222],{},"Today",", ",[170,225,226],{},"7d",[170,228,229],{},"30d"," or ",[170,232,233],{},"Custom",". Days start at midnight in the Europe\u002FMadrid time zone.",[197,236,237,242],{},[212,238,239],{},[170,240,241],{},"Comparison",[212,243,244,245,248,249,189],{},"The console compares each figure with the same figure for the period of the same length immediately before. ",[170,246,247],{},"Change %"," is the change divided by the earlier value. If the earlier value is 0, the console shows no ",[170,250,247],{},[197,252,253,258],{},[212,254,255],{},[170,256,257],{},"Request",[212,259,260],{},"One API request from an application to the Gateway. Failed and blocked requests count.",[197,262,263,268],{},[212,264,265],{},[170,266,267],{},"Provider call",[212,269,270],{},"One attempt to send a request to a provider. A retry or a fallback adds a call. A request that a guardrail blocks has no call.",[197,272,273,278],{},[212,274,275],{},[170,276,277],{},"Empty values",[212,279,280],{},"When a rate, an average or a percentile has no data, the console shows no value. It does not show 0.",[282,283,285],"h3",{"id":284},"charts-and-tables","Charts and tables",[166,287,288],{},"Each section shows one metric at a time. You can show it as a chart or as a table.",[290,291,292,321],"ul",{},[293,294,295,298,299,223,301,223,304,223,307,223,310,230,313,316,317,320],"li",{},[170,296,297],{},"Grouping."," You can split a metric by ",[170,300,94],{},[170,302,303],{},"Project",[170,305,306],{},"Provider",[170,308,309],{},"Model",[170,311,312],{},"Tool",[170,314,315],{},"Reason",". ",[170,318,319],{},"Total"," shows the metric with no split. Each metric offers only the groupings that apply to it.",[293,322,323,326],{},[170,324,325],{},"Bucket."," The bucket is the time step of the chart: hour, day, week or month. The console picks the smallest step that gives 62 points or fewer. Weeks start on Monday. The first and the last bucket can be partial.",[282,328,330],{"id":329},"when-the-rows-and-the-total-differ","When the rows and the total differ",[332,333,336,337,230,339,341],"callout",{"color":334,"icon":335},"warning","i-lucide-triangle-alert","Grouped by ",[170,338,306],{},[170,340,309],{},", the rows of some metrics do not add up to the total.",[166,343,344,345,223,348,223,351,354,355,358],{},"It affects ",[170,346,347],{},"Inference requests",[170,349,350],{},"Successful requests",[170,352,353],{},"Errored requests"," and ",[170,356,357],{},"Failed requests",". Average cost per request and Success rate divide by these counts, so their Provider and Model rows are affected too. With every other grouping, the rows of a count, a token sum or Spend add up to the total.",[166,360,361],{},"Two causes change the sum of the rows:",[290,363,364,374],{},[293,365,366,369,370,373],{},[170,367,368],{},"Fallback or retry."," One request reaches more than one provider or model, and counts once under each. The rows add up to ",[170,371,372],{},"more"," than the total.",[293,375,376,379,380,373],{},[170,377,378],{},"Blocked before any provider call."," The request counts in the total and in no row. The rows add up to ",[170,381,382],{},"less",[166,384,385],{},"Both causes can occur in the same period and cancel out. The rows then match the total, but they still count some requests twice and leave others out.",[387,388],"request-calls",{":requests":389,"caption":390},"[{\"label\":\"Request A\",\"calls\":\"1 provider call\",\"steps\":[{\"label\":\"OpenAI\",\"outcome\":\"success\"}]},{\"label\":\"Request B\",\"calls\":\"2 provider calls\",\"steps\":[{\"label\":\"OpenAI\",\"outcome\":\"error\"},{\"label\":\"Anthropic\",\"outcome\":\"success\",\"via\":\"fallback\"}]},{\"label\":\"Request C\",\"calls\":\"0 provider calls\",\"steps\":[{\"label\":\"Guardrail blocks it\",\"outcome\":\"blocked\"}]}]","Three requests, and the provider calls each one makes",[392,393],"overlap-rows",{":cases":394,"caption":395,"grouping":306,"metric":347},"[{\"title\":\"Fallback\",\"note\":\"Request B counts under OpenAI and under Anthropic.\",\"total\":1,\"rows\":[{\"label\":\"OpenAI\",\"value\":1},{\"label\":\"Anthropic\",\"value\":1}]},{\"title\":\"Block\",\"note\":\"Request A counts under OpenAI. Request C counts in no row.\",\"total\":2,\"rows\":[{\"label\":\"OpenAI\",\"value\":1}]},{\"title\":\"Both\",\"note\":\"Requests A, B and C in one period.\",\"total\":3,\"rows\":[{\"label\":\"OpenAI\",\"value\":2},{\"label\":\"Anthropic\",\"value\":1}]}]","In the third case the rows match the total, but request B counts twice and request C counts in no row.",[178,397,399],{"id":398},"usage","Usage",[191,401,402,421],{},[194,403,404],{},[197,405,406,409,412,415,418],{},[200,407,408],{},"Metric",[200,410,411],{},"Definition",[200,413,414],{},"How it's calculated",[200,416,417],{},"Unit",[200,419,420],{},"Where it appears",[207,422,423,446,465,482,500,517],{},[197,424,425,429,432,440,443],{},[212,426,427],{},[170,428,347],{},[212,430,431],{},"Inference requests from applications to the Gateway. Failed and blocked requests count.",[212,433,434,435,189],{},"Count of requests. See ",[436,437,439],"a",{"href":438},"#when-the-rows-and-the-total-differ","when the rows and the total differ",[212,441,442],{},"Count",[212,444,445],{},"Dashboard: summary card and chart. Projects list, for the last 30 days.",[197,447,448,453,456,459,462],{},[212,449,450],{},[170,451,452],{},"Input tokens",[212,454,455],{},"Prompt tokens sent to providers.",[212,457,458],{},"Sum of input tokens of every provider call, retries included. The sum includes tokens that the provider reads from or writes to its prompt cache.",[212,460,461],{},"Tokens",[212,463,464],{},"Dashboard: summary card and chart",[197,466,467,472,475,478,480],{},[212,468,469],{},[170,470,471],{},"Output tokens",[212,473,474],{},"Tokens that providers generated.",[212,476,477],{},"Sum of output tokens of every provider call. The sum includes reasoning tokens.",[212,479,461],{},[212,481,464],{},[197,483,484,489,492,495,497],{},[212,485,486],{},[170,487,488],{},"Total tokens",[212,490,491],{},"All tokens of every provider call.",[212,493,494],{},"Input tokens + output tokens.",[212,496,461],{},[212,498,499],{},"Dashboard: chart",[197,501,502,507,510,513,515],{},[212,503,504],{},[170,505,506],{},"Reasoning tokens",[212,508,509],{},"Output tokens that a model used to think before it answered.",[212,511,512],{},"Sum of reasoning tokens that providers report. Some providers report 0.",[212,514,461],{},[212,516,499],{},[197,518,519,524,527,530,533],{},[212,520,521],{},[170,522,523],{},"Cache hit rate",[212,525,526],{},"Share of input tokens that providers read from their prompt cache.",[212,528,529],{},"Cached input tokens ÷ input tokens.",[212,531,532],{},"%",[212,534,464],{},[178,536,538],{"id":537},"cost","Cost",[191,540,541,555],{},[194,542,543],{},[197,544,545,547,549,551,553],{},[200,546,408],{},[200,548,411],{},[200,550,414],{},[200,552,417],{},[200,554,420],{},[207,556,557,578,596,615],{},[197,558,559,564,570,573,576],{},[212,560,561],{},[170,562,563],{},"Spend",[212,565,566,567,189],{},"What the provider calls cost at ",[170,568,569],{},"list price",[212,571,572],{},"Sum of the price of every priced provider call, retries included. An unpriced call adds 0.",[212,574,575],{},"USD",[212,577,464],{},[197,579,580,585,588,591,593],{},[212,581,582],{},[170,583,584],{},"Average spend per day",[212,586,587],{},"Spend spread evenly over the period.",[212,589,590],{},"Spend ÷ days in the period. A day with no spend still counts.",[212,592,575],{},[212,594,595],{},"Dashboard: summary card",[197,597,598,603,606,611,613],{},[212,599,600],{},[170,601,602],{},"Average cost per request",[212,604,605],{},"What one request costs on average.",[212,607,608,609,189],{},"Spend ÷ requests. Requests that cost nothing count. By provider or model, see ",[436,610,439],{"href":438},[212,612,575],{},[212,614,464],{},[197,616,617,622,625,632,634],{},[212,618,619],{},[170,620,621],{},"Blended price per million tokens",[212,623,624],{},"The average price of one million tokens across all models.",[212,626,627,628,631],{},"Spend × 1,000,000 ÷ tokens of ",[170,629,630],{},"priced"," calls. The divisor leaves out unpriced calls.",[212,633,575],{},[212,635,464],{},[332,637,639,642],{"color":638,"icon":335},"info",[170,640,641],{},"USD is the standard currency."," Costs are reported in USD because this is the currency providers use to price their models.",[282,644,646],{"id":645},"where-prices-come-from","Where prices come from",[166,648,649],{},"Optiak prices each provider call when the request runs. It uses the price of the model that served the call:",[290,651,652,658],{},[293,653,654,657],{},[170,655,656],{},"Built-in models"," use the provider's list price. The price includes the cache read and cache write rates, the long-context rates, and the priority and flex tiers where the provider has them.",[293,659,660,663,664,667,668,189],{},[170,661,662],{},"Custom provider models"," use the prices that an Organization Admin sets under ",[170,665,666],{},"Pricing (optional)",": input and output, plus cache read and cache write if set. A cache price left blank uses the input price. A price of 0 makes cached tokens free. See ",[436,669,671],{"href":670},"\u002Fadministration\u002Fauthentication\u002Fconnect-model-apis#custom-providers","Custom providers",[332,673,674,677],{"color":334,"icon":335},[170,675,676],{},"Spend is an estimate."," It uses list prices, so it is not your provider invoice. Spend does not include discounts, credits or committed-use prices. A call to a model with no price is unpriced. An unpriced call adds 0 to Spend. So when some models have no price, Spend is lower than the list-price cost.",[282,679,681],{"id":680},"how-a-change-splits-into-volume-mix-and-rate","How a change splits into volume, mix and rate",[166,683,684,685,688],{},"In the ",[170,686,687],{},"Table"," view, each row shows its part of the group's change over the previous period, and splits that change into key drivers:",[191,690,691,701],{},[194,692,693],{},[197,694,695,698],{},[200,696,697],{},"Column",[200,699,700],{},"Meaning",[207,702,703,719,729,739,749,759],{},[197,704,705,710],{},[212,706,707],{},[170,708,709],{},"Share of change",[212,711,712,713,715,716,189],{},"How much of the group's whole change this row caused, in percentage points. The rows add up to the group's ",[170,714,247],{},", except when ",[436,717,718],{"href":438},"the rows and the total differ",[197,720,721,726],{},[212,722,723],{},[170,724,725],{},"Volume",[212,727,728],{},"The group's total tokens changed. This row kept its share.",[197,730,731,736],{},[212,732,733],{},[170,734,735],{},"Mix",[212,737,738],{},"Rows gained or lost share. The group's total stayed the same.",[197,740,741,746],{},[212,742,743],{},[170,744,745],{},"Rate",[212,747,748],{},"The row's own price for each token changed.",[197,750,751,756],{},[212,752,753],{},[170,754,755],{},"Entered",[212,757,758],{},"The row was at 0 in the earlier period. This column holds its whole change.",[197,760,761,766],{},[212,762,763],{},[170,764,765],{},"Left",[212,767,768],{},"The row is at 0 in the later period. This column holds its whole change.",[166,770,771],{},"Volume-Mix-Rate is a standard decomposition used in financial and operational analytics to decompose variance in costs into distinct drivers: overall quantity changes (Volume), changes in product or customer composition (Mix), and changes in unit costs (Rate). Here these concepts are specialized to the cost of requests to model providers.",[773,774,777,785,791,969],"worked-example",{"description":775,"title":776},"A change of +$700, split by provider, with every number shown.","Worked example: why Spend changed",[778,779],"change-split",{":before":780,":parts":781,"caption":782,"prefix":783,"title":784},"1000","[{\"label\":\"Volume\",\"value\":659,\"description\":\"The continuing providers used more tokens in total.\"},{\"label\":\"Mix\",\"value\":-89,\"description\":\"Tokens moved from OpenAI to the cheaper Anthropic.\"},{\"label\":\"Rate\",\"value\":80,\"description\":\"Anthropic's price for each token went up.\"},{\"label\":\"Entered\",\"value\":100,\"description\":\"Mistral had no spend in the earlier period.\"},{\"label\":\"Left\",\"value\":-50,\"description\":\"Cohere has no spend in the later period.\"}]","Example numbers. Spend went from $1,000 to $1,700. The five parts add up to the +$700 change.","$","Why Spend changed, by provider",[166,786,787,788,790],{},"Spend grouped by ",[170,789,306],{},". M means million tokens. Amounts are rounded to whole dollars.",[191,792,793,816],{},[194,794,795],{},[197,796,797,799,801,804,806,808,810,812,814],{},[200,798,306],{},[200,800,461],{},[200,802,803],{},"Price per 1M tokens",[200,805,563],{},[200,807,725],{},[200,809,735],{},[200,811,745],{},[200,813,755],{},[200,815,765],{},[207,817,818,847,876,903,929],{},[197,819,820,825,828,831,834,837,840,843,845],{},[212,821,822],{},[170,823,824],{},"OpenAI",[212,826,827],{},"50M → 60M",[212,829,830],{},"$12 → $12",[212,832,833],{},"$600 → $720",[212,835,836],{},"+$358",[212,838,839],{},"−$238",[212,841,842],{},"$0",[212,844],{},[212,846],{},[197,848,849,854,857,860,863,866,869,872,874],{},[212,850,851],{},[170,852,853],{},"Anthropic",[212,855,856],{},"50M → 110M",[212,858,859],{},"$7 → $8",[212,861,862],{},"$350 → $880",[212,864,865],{},"+$301",[212,867,868],{},"+$149",[212,870,871],{},"+$80",[212,873],{},[212,875],{},[197,877,878,883,886,889,892,894,896,898,901],{},[212,879,880],{},[170,881,882],{},"Mistral",[212,884,885],{},"0 → 20M",[212,887,888],{},"$5",[212,890,891],{},"$0 → $100",[212,893],{},[212,895],{},[212,897],{},[212,899,900],{},"+$100",[212,902],{},[197,904,905,910,913,915,918,920,922,924,926],{},[212,906,907],{},[170,908,909],{},"Cohere",[212,911,912],{},"10M → 0",[212,914,888],{},[212,916,917],{},"$50 → $0",[212,919],{},[212,921],{},[212,923],{},[212,925],{},[212,927,928],{},"−$50",[197,930,931,935,940,942,947,952,957,961,965],{},[212,932,933],{},[170,934,319],{},[212,936,937],{},[170,938,939],{},"110M → 190M",[212,941],{},[212,943,944],{},[170,945,946],{},"$1,000 → $1,700",[212,948,949],{},[170,950,951],{},"+$659",[212,953,954],{},[170,955,956],{},"−$89",[212,958,959],{},[170,960,871],{},[212,962,963],{},[170,964,900],{},[212,966,967],{},[170,968,928],{},[290,970,971,977,983,989,995],{},[293,972,973,976],{},[170,974,975],{},"Volume +$659."," OpenAI and Anthropic used 100M tokens before and 170M after. That growth, at each provider's share and price, adds $659.",[293,978,979,982],{},[170,980,981],{},"Mix −$89."," OpenAI's share of tokens fell from 50% to 35%, and Anthropic's grew from 50% to 65%. Anthropic is cheaper, so the move lowers Spend. OpenAI loses $238, and Anthropic gains $149.",[293,984,985,988],{},[170,986,987],{},"Rate +$80."," Anthropic's price went from $7 to $8 per 1M tokens, on 80M tokens on average. OpenAI's price did not change, so its Rate is $0.",[293,990,991,994],{},[170,992,993],{},"Entered +$100."," Mistral had no spend before, so all of its $100 is Entered.",[293,996,997,1000],{},[170,998,999],{},"Left −$50."," Cohere has no spend now, so all of its $50 is Left.",[166,1002,1003],{},"Volume, Mix, Rate, Entered and Left add up to the row's Share of change.",[166,1005,1006],{},"Average cost per request and Blended price per million tokens are rates, so they have no Volume column. Their change splits into these parts:",[290,1008,1009,1014,1019],{},[293,1010,1011,1013],{},[170,1012,745],{},": the row's own rate changed. Its weight stayed the same.",[293,1015,1016,1018],{},[170,1017,735],{},": weight moved from some rows to other rows. The rates stayed the same.",[293,1020,1021,354,1023,1025],{},[170,1022,755],{},[170,1024,765],{},": the row had no weight in one of the two periods. Optiak calculates this part from the difference between the row's rate and the group's average rate.",[166,1027,1028],{},"The weight of a row is the divisor of its rate. It is requests for Average cost per request, and priced tokens for Blended price per million tokens.",[773,1030,1033,1041,1047,1206],{"description":1031,"title":1032},"A change of +$0.06, split by provider. A new provider can lower a rate.","Worked example: why Average cost per request changed",[778,1034],{":before":1035,":parts":1036,"caption":1037,"prefix":783,"title":1038,":axisFrom":1039,":decimals":1040},"0.97","[{\"label\":\"Rate\",\"value\":0.09,\"description\":\"Anthropic's cost per request went up.\"},{\"label\":\"Mix\",\"value\":-0.03,\"description\":\"Requests moved toward providers that cost less than the average.\"},{\"label\":\"Entered\",\"value\":-0.05,\"description\":\"Mistral joined, and it costs less than the average.\"},{\"label\":\"Left\",\"value\":0.05,\"description\":\"Cohere left, and it cost less than the average.\"}]","Example numbers. The average went from $0.97 to $1.03. The four parts add up to the +$0.06 change. The axis starts at $0.80.","Why Average cost per request changed, by provider","0.8","2",[166,1042,1043,1044,1046],{},"Average cost per request grouped by ",[170,1045,306],{},". The weight of each row is its share of requests. The group's average cost per request over the two periods is $1.00.",[191,1048,1049,1072],{},[194,1050,1051],{},[197,1052,1053,1055,1058,1061,1064,1066,1068,1070],{},[200,1054,306],{},[200,1056,1057],{},"Requests",[200,1059,1060],{},"Cost per request",[200,1062,1063],{},"Share of requests",[200,1065,745],{},[200,1067,735],{},[200,1069,755],{},[200,1071,765],{},[207,1073,1074,1099,1124,1148,1171],{},[197,1075,1076,1080,1083,1086,1089,1092,1095,1097],{},[212,1077,1078],{},[170,1079,824],{},[212,1081,1082],{},"50 → 80",[212,1084,1085],{},"$1.20 → $1.20",[212,1087,1088],{},"50% → 40%",[212,1090,1091],{},"$0.00",[212,1093,1094],{},"−$0.02",[212,1096],{},[212,1098],{},[197,1100,1101,1105,1108,1111,1114,1117,1120,1122],{},[212,1102,1103],{},[170,1104,853],{},[212,1106,1107],{},"40 → 100",[212,1109,1110],{},"$0.80 → $1.00",[212,1112,1113],{},"40% → 50%",[212,1115,1116],{},"+$0.09",[212,1118,1119],{},"−$0.01",[212,1121],{},[212,1123],{},[197,1125,1126,1130,1133,1136,1139,1141,1143,1146],{},[212,1127,1128],{},[170,1129,882],{},[212,1131,1132],{},"0 → 20",[212,1134,1135],{},"$0.50",[212,1137,1138],{},"0% → 10%",[212,1140],{},[212,1142],{},[212,1144,1145],{},"−$0.05",[212,1147],{},[197,1149,1150,1154,1157,1159,1162,1164,1166,1168],{},[212,1151,1152],{},[170,1153,909],{},[212,1155,1156],{},"10 → 0",[212,1158,1135],{},[212,1160,1161],{},"10% → 0%",[212,1163],{},[212,1165],{},[212,1167],{},[212,1169,1170],{},"+$0.05",[197,1172,1173,1177,1182,1187,1189,1193,1198,1202],{},[212,1174,1175],{},[170,1176,319],{},[212,1178,1179],{},[170,1180,1181],{},"100 → 200",[212,1183,1184],{},[170,1185,1186],{},"$0.97 → $1.03",[212,1188],{},[212,1190,1191],{},[170,1192,1116],{},[212,1194,1195],{},[170,1196,1197],{},"−$0.03",[212,1199,1200],{},[170,1201,1145],{},[212,1203,1204],{},[170,1205,1170],{},[290,1207,1208,1214,1220,1226],{},[293,1209,1210,1213],{},[170,1211,1212],{},"Rate +$0.09."," Anthropic's cost per request went from $0.80 to $1.00. Its average share is 45%, so the average rises by $0.09. OpenAI's cost did not change, so its Rate is $0.00.",[293,1215,1216,1219],{},[170,1217,1218],{},"Mix −$0.03."," OpenAI costs more than the $1.00 average, and its share fell from 50% to 40%: −$0.02. Anthropic's mean cost, $0.90, is below the average, and its share grew: −$0.01.",[293,1221,1222,1225],{},[170,1223,1224],{},"Entered −$0.05."," Mistral costs $0.50, half the average, and now has 10% of requests. A cheap new row pulls the average down.",[293,1227,1228,1231],{},[170,1229,1230],{},"Left +$0.05."," Cohere cost $0.50 and had 10% of requests. A cheap row that leaves pushes the average up.",[166,1233,1234],{},"The other metrics show only Share of change. Percentiles show no split.",[178,1236,1238],{"id":1237},"reliability","Reliability",[166,1240,1241,1242,230,1244,1246,1247,189],{},"When you group by ",[170,1243,306],{},[170,1245,309],{},", the counts below are provider calls, not requests. See ",[436,1248,439],{"href":438},[191,1250,1251,1265],{},[194,1252,1253],{},[197,1254,1255,1257,1259,1261,1263],{},[200,1256,408],{},[200,1258,411],{},[200,1260,414],{},[200,1262,417],{},[200,1264,420],{},[207,1266,1267,1283,1299,1319,1340,1358,1375,1392,1409],{},[197,1268,1269,1273,1276,1279,1281],{},[212,1270,1271],{},[170,1272,350],{},[212,1274,1275],{},"Requests that succeeded.",[212,1277,1278],{},"Count of successful requests. By provider or model: successful calls.",[212,1280,442],{},[212,1282,464],{},[197,1284,1285,1289,1292,1295,1297],{},[212,1286,1287],{},[170,1288,353],{},[212,1290,1291],{},"Requests that ended in an error. Blocked requests are not errors.",[212,1293,1294],{},"Count of errored requests. By provider or model: errored calls.",[212,1296,442],{},[212,1298,499],{},[197,1300,1301,1305,1308,1315,1317],{},[212,1302,1303],{},[170,1304,357],{},[212,1306,1307],{},"Requests that did not succeed.",[212,1309,1310,1311,1314],{},"Errored requests + ",[170,1312,1313],{},"blocked requests",". By provider or model: errored calls.",[212,1316,442],{},[212,1318,464],{},[197,1320,1321,1326,1329,1336,1338],{},[212,1322,1323],{},[170,1324,1325],{},"Success rate",[212,1327,1328],{},"Share of requests that succeeded, without the requests that a guardrail blocked. A block shows that the policy works. It is not a failure.",[212,1330,1331,1332,1335],{},"Successful requests ÷ (successful requests + errored requests). Blocked requests are ",[170,1333,1334],{},"not"," in the divisor. By provider or model: over calls.",[212,1337,532],{},[212,1339,464],{},[197,1341,1342,1347,1350,1353,1356],{},[212,1343,1344],{},[170,1345,1346],{},"End-to-end latency p50",[212,1348,1349],{},"The median time from request to response. Half of the requests are faster.",[212,1351,1352],{},"50th percentile of request duration. The percentile leaves out requests with no measured time.",[212,1354,1355],{},"ms",[212,1357,499],{},[197,1359,1360,1365,1368,1371,1373],{},[212,1361,1362],{},[170,1363,1364],{},"End-to-end latency p95",[212,1366,1367],{},"The time that 95% of requests stay under.",[212,1369,1370],{},"95th percentile of request duration. The percentile leaves out requests with no measured time.",[212,1372,1355],{},[212,1374,464],{},[197,1376,1377,1382,1385,1388,1390],{},[212,1378,1379],{},[170,1380,1381],{},"Provider latency p50",[212,1383,1384],{},"The median time a request spent in provider calls.",[212,1386,1387],{},"50th percentile of the provider time per request, retries added. The percentile leaves out requests with no provider call.",[212,1389,1355],{},[212,1391,499],{},[197,1393,1394,1399,1402,1405,1407],{},[212,1395,1396],{},[170,1397,1398],{},"Provider latency p95",[212,1400,1401],{},"The provider time that 95% of requests stay under.",[212,1403,1404],{},"95th percentile of the provider time per request, retries added. The percentile leaves out requests with no provider call.",[212,1406,1355],{},[212,1408,499],{},[197,1410,1411,1416,1419,1422,1424],{},[212,1412,1413],{},[170,1414,1415],{},"Top failure reasons",[212,1417,1418],{},"The most frequent reasons why requests failed.",[212,1420,1421],{},"Count of requests with a failure reason, grouped by reason. Blocked requests count.",[212,1423,442],{},[212,1425,1426],{},"Dashboard: list",[178,1428,1430],{"id":1429},"governance-security","Governance & Security",[166,1432,1433,1434,1436],{},"These metrics come from ",[436,1435,58],{"href":59},". One request can run several checks.",[191,1438,1439,1453],{},[194,1440,1441],{},[197,1442,1443,1445,1447,1449,1451],{},[200,1444,408],{},[200,1446,411],{},[200,1448,414],{},[200,1450,417],{},[200,1452,420],{},[207,1454,1455,1472,1489,1510,1527,1547],{},[197,1456,1457,1462,1465,1468,1470],{},[212,1458,1459],{},[170,1460,1461],{},"Total evaluations",[212,1463,1464],{},"Guardrail checks that ran.",[212,1466,1467],{},"Count of checks.",[212,1469,442],{},[212,1471,464],{},[197,1473,1474,1479,1482,1485,1487],{},[212,1475,1476],{},[170,1477,1478],{},"Blocked requests",[212,1480,1481],{},"Requests that a guardrail stopped.",[212,1483,1484],{},"Count of blocked requests. A request that several checks block counts once.",[212,1486,442],{},[212,1488,499],{},[197,1490,1491,1496,1499,1506,1508],{},[212,1492,1493],{},[170,1494,1495],{},"Blocked rate",[212,1497,1498],{},"Blocked requests for each guardrail check that ran.",[212,1500,1501,1502,1505],{},"Blocked requests ÷ checks that ran. The divisor is ",[170,1503,1504],{},"checks",", not requests.",[212,1507,532],{},[212,1509,464],{},[197,1511,1512,1517,1520,1523,1525],{},[212,1513,1514],{},[170,1515,1516],{},"Guardrail added latency",[212,1518,1519],{},"The average time of one guardrail check.",[212,1521,1522],{},"Total check time ÷ checks. A check with no measured time counts as 0 ms.",[212,1524,1355],{},[212,1526,464],{},[197,1528,1529,1534,1537,1543,1545],{},[212,1530,1531],{},[170,1532,1533],{},"Audit detection rate",[212,1535,1536],{},"Share of checks that found a problem.",[212,1538,1539,1540,189],{},"Checks that detected, blocked, redacted or replaced ÷ checks that did ",[170,1541,1542],{},"not end in an error",[212,1544,532],{},[212,1546,464],{},[197,1548,1549,1554,1557,1560,1562],{},[212,1550,1551],{},[170,1552,1553],{},"Violations by application",[212,1555,1556],{},"Checks that found a problem, for each application.",[212,1558,1559],{},"Count of checks that detected, blocked, redacted or replaced, grouped by application.",[212,1561,442],{},[212,1563,1564],{},"Dashboard: table",[178,1566,48],{"id":1567},"data-integrations",[166,1569,1570,1571,1573],{},"These metrics come from the retrieval engine in ",[436,1572,48],{"href":49},". One retrieval is one search. One request can run several retrievals.",[191,1575,1576,1590],{},[194,1577,1578],{},[197,1579,1580,1582,1584,1586,1588],{},[200,1581,408],{},[200,1583,411],{},[200,1585,414],{},[200,1587,417],{},[200,1589,420],{},[207,1591,1592,1609,1630,1647],{},[197,1593,1594,1599,1602,1605,1607],{},[212,1595,1596],{},[170,1597,1598],{},"Average chunks per request",[212,1600,1601],{},"Documents that retrieval added to one request, on average.",[212,1603,1604],{},"Documents retrieved ÷ requests that used retrieval.",[212,1606,442],{},[212,1608,464],{},[197,1610,1611,1616,1619,1626,1628],{},[212,1612,1613],{},[170,1614,1615],{},"Retrieval failures",[212,1617,1618],{},"Requests where retrieval failed.",[212,1620,1621,1622,1625],{},"Count of requests with a failed retrieval. A request counts ",[170,1623,1624],{},"once",". No value if no retrieval ran.",[212,1627,442],{},[212,1629,464],{},[197,1631,1632,1637,1640,1643,1645],{},[212,1633,1634],{},[170,1635,1636],{},"Tokens contributed by retrieved chunks",[212,1638,1639],{},"Tokens of the documents that retrieval added to prompts.",[212,1641,1642],{},"Sum of document tokens. No value if no retrieval ran.",[212,1644,461],{},[212,1646,464],{},[197,1648,1649,1654,1657,1660,1662],{},[212,1650,1651],{},[170,1652,1653],{},"Retrieval added latency",[212,1655,1656],{},"The average time of one retrieval.",[212,1658,1659],{},"Total retrieval time ÷ retrievals. A retrieval with no measured time counts as 0 ms.",[212,1661,1355],{},[212,1663,464],{},[178,1665,53],{"id":1666},"tools",[166,1668,1669,1670,1672,1673,1677,1678,1681],{},"These metrics come from tool calls through the MCP Gateway. See ",[436,1671,53],{"href":54},". A tool call is an MCP ",[1674,1675,1676],"code",{},"tools\u002Fcall"," request. A tool list request is an MCP ",[1674,1679,1680],{},"tools\u002Flist"," request.",[191,1683,1684,1698],{},[194,1685,1686],{},[197,1687,1688,1690,1692,1694,1696],{},[200,1689,408],{},[200,1691,411],{},[200,1693,414],{},[200,1695,417],{},[200,1697,420],{},[207,1699,1700,1720,1739,1756,1773,1790,1807,1824],{},[197,1701,1702,1707,1710,1716,1718],{},[212,1703,1704],{},[170,1705,1706],{},"MCP tool calls",[212,1708,1709],{},"Requests to run a tool.",[212,1711,1712,1713,1715],{},"Count of ",[1674,1714,1676],{}," requests. The count includes calls that ended in an error before the tool ran.",[212,1717,442],{},[212,1719,464],{},[197,1721,1722,1727,1730,1735,1737],{},[212,1723,1724],{},[170,1725,1726],{},"MCP tool list requests",[212,1728,1729],{},"Requests that ask a server for its tools. They run no tool.",[212,1731,1712,1732,1734],{},[1674,1733,1680],{}," requests.",[212,1736,442],{},[212,1738,499],{},[197,1740,1741,1746,1749,1752,1754],{},[212,1742,1743],{},[170,1744,1745],{},"Successful MCP tool calls",[212,1747,1748],{},"Tool calls that completed without an error.",[212,1750,1751],{},"Count of successful tool calls.",[212,1753,442],{},[212,1755,499],{},[197,1757,1758,1763,1766,1769,1771],{},[212,1759,1760],{},[170,1761,1762],{},"Failed MCP tool calls",[212,1764,1765],{},"Tool calls that ended in an error.",[212,1767,1768],{},"Count of failed tool calls. The count includes calls to an unknown tool and calls that went past the time limit.",[212,1770,442],{},[212,1772,464],{},[197,1774,1775,1780,1783,1786,1788],{},[212,1776,1777],{},[170,1778,1779],{},"MCP tool call success rate",[212,1781,1782],{},"Share of tool calls that succeeded.",[212,1784,1785],{},"Successful MCP tool calls ÷ MCP tool calls. The divisor leaves out tool list requests.",[212,1787,532],{},[212,1789,464],{},[197,1791,1792,1797,1800,1803,1805],{},[212,1793,1794],{},[170,1795,1796],{},"MCP tool latency p50",[212,1798,1799],{},"The median time a tool ran.",[212,1801,1802],{},"50th percentile of tool run time per call. The percentile leaves out calls where no tool ran.",[212,1804,1355],{},[212,1806,499],{},[197,1808,1809,1814,1817,1820,1822],{},[212,1810,1811],{},[170,1812,1813],{},"MCP tool latency p95",[212,1815,1816],{},"The tool run time that 95% of calls stay under.",[212,1818,1819],{},"95th percentile of tool run time per call. The percentile leaves out calls where no tool ran.",[212,1821,1355],{},[212,1823,464],{},[197,1825,1826,1831,1834,1837,1839],{},[212,1827,1828],{},[170,1829,1830],{},"Top MCP failure categories",[212,1832,1833],{},"The most frequent causes of failed MCP tool calls.",[212,1835,1836],{},"Count of failed MCP tool calls, grouped by the tool's failure category. A call with no category counts under its request failure reason.",[212,1838,442],{},[212,1840,1426],{},[178,1842,1844],{"id":1843},"application-overview","Application overview",[166,1846,1847,1848,1850,1851,1853,1854,230,1856,1858],{},"An application's ",[170,1849,28],{}," shows the Dashboard sections for that application. The Overview uses the metric definitions on this page, with the data of that application only. Where a table says ",[170,1852,172],{},", the Overview shows the same card or chart. The Overview has no ",[170,1855,94],{},[170,1857,303],{}," grouping.",[178,1860,1862],{"id":1861},"related","Related",[290,1864,1865,1870],{},[293,1866,1867,1869],{},[436,1868,94],{"href":95},": the Overview and Requests tabs of one application.",[293,1871,1872,1874],{},[436,1873,145],{"href":146},": spend limits for projects and applications.",{"title":1876,"searchDepth":1877,"depth":1877,"links":1878},"",2,[1879,1884,1885,1889,1890,1891,1892,1893,1894],{"id":180,"depth":1877,"text":181,"children":1880},[1881,1883],{"id":284,"depth":1882,"text":285},3,{"id":329,"depth":1882,"text":330},{"id":398,"depth":1877,"text":399},{"id":537,"depth":1877,"text":538,"children":1886},[1887,1888],{"id":645,"depth":1882,"text":646},{"id":680,"depth":1882,"text":681},{"id":1237,"depth":1877,"text":1238},{"id":1429,"depth":1877,"text":1430},{"id":1567,"depth":1877,"text":48},{"id":1666,"depth":1877,"text":53},{"id":1843,"depth":1877,"text":1844},{"id":1861,"depth":1877,"text":1862},"The definition of every metric in the console, and how Optiak calculates it.","md",null,{},{"icon":102},{"title":99,"description":1895},"dz02PX1Hy6KeK3fOZlzblUv-EKKcAOHQiG7LO8BM8qE",[1903,1905],{"title":94,"path":95,"stem":96,"description":1904,"icon":97,"children":-1},"The traffic, cost and individual requests of one application.",{"title":110,"path":111,"stem":112,"description":1906,"icon":113,"children":-1},"How projects group applications and how you manage them.",1795006120970]