{"id":2778,"date":"2026-08-12T15:30:00","date_gmt":"2026-08-12T15:30:00","guid":{"rendered":"https:\/\/dmarketertayeeb.com\/blog\/?p=2778"},"modified":"2026-08-12T16:05:11","modified_gmt":"2026-08-12T16:05:11","slug":"ai-agent-cost-per-accepted-result-calculator","status":"publish","type":"post","link":"https:\/\/dmarketertayeeb.com\/blog\/ai-agent-cost-per-accepted-result-calculator\/","title":{"rendered":"AI Agent Cost per Accepted Result Calculator and Evaluation Template"},"content":{"rendered":"\n\n\n<p><strong>Cost per accepted result<\/strong> is the total cost of every run, retry, tool call and human review divided by the number of outputs that pass a defined acceptance rubric. It is more useful than price per million tokens because a cheap model can become expensive when it fails often, and an expensive model can be economical when it reduces retries and correction time.<\/p>\n\n\n\n\n\n\n\n<p>The calculator below turns DMT\u2019s <a href=\"https:\/\/dmarketertayeeb.com\/blog\/gpt-5-6-luna-terra-sol-cost-per-accepted-result\/\">accepted-result methodology<\/a> into a working tool. It uses editable price fields and August 11, 2026 OpenAI GPT-5.6 presets. Always check the <a href=\"https:\/\/developers.openai.com\/api\/docs\/pricing\">current official pricing page<\/a> before a procurement or budget decision.<\/p>\n\n\n\n\n\n\n\n<figure class=\"wp-block-image size-full\"><a href=\"https:\/\/developers.openai.com\/api\/docs\/pricing\"><img loading=\"lazy\" decoding=\"async\" width=\"900\" height=\"191\" src=\"https:\/\/dmarketertayeeb.com\/blog\/wp-content\/uploads\/2026\/08\/openai-gpt-5-6-pricing-table-2026-08-11.png\" alt=\"OpenAI GPT-5.6 Sol, Terra and Luna token pricing table\" class=\"wp-image-2777\" srcset=\"https:\/\/dmarketertayeeb.com\/blog\/wp-content\/uploads\/2026\/08\/openai-gpt-5-6-pricing-table-2026-08-11.png 900w, https:\/\/dmarketertayeeb.com\/blog\/wp-content\/uploads\/2026\/08\/openai-gpt-5-6-pricing-table-2026-08-11-300x64.png 300w, https:\/\/dmarketertayeeb.com\/blog\/wp-content\/uploads\/2026\/08\/openai-gpt-5-6-pricing-table-2026-08-11-768x163.png 768w\" sizes=\"auto, (max-width: 900px) 100vw, 900px\" \/><\/a><figcaption class=\"wp-element-caption\">Official GPT-5.6 token pricing used for the editable calculator presets. Source: OpenAI; captured August 11, 2026. Always verify current rates.<\/figcaption><\/figure>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">AI agent cost calculator<\/h2>\n\n\n\n\n\n<div id=\"dmt-agent-cost-calculator\" style=\"border:1px solid #d7dce1;border-radius:12px;padding:20px;margin:24px 0;background:#f8fafc;color:#111827\">\n  \n\n<p><strong>All prices are editable.<\/strong> Token rates are USD per 1 million tokens. Enter totals across the evaluation batch, not a single successful run.<\/p>\n\n\n  <div style=\"display:grid;grid-template-columns:repeat(auto-fit,minmax(210px,1fr));gap:14px\">\n    <label>Price preset<br><select id=\"dmt-preset\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><option value=\"luna\">GPT-5.6 Luna short<\/option><option value=\"terra\">GPT-5.6 Terra short<\/option><option value=\"sol\">GPT-5.6 Sol short<\/option><option value=\"custom\">Custom<\/option><\/select><\/label>\n    <label>Total attempted runs<br><input id=\"dmt-runs\" type=\"number\" min=\"1\" step=\"1\" value=\"100\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Accepted results<br><input id=\"dmt-accepted\" type=\"number\" min=\"0\" step=\"1\" value=\"80\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Input tokens (total)<br><input id=\"dmt-input-tokens\" type=\"number\" min=\"0\" step=\"1\" value=\"5000000\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Cached input tokens (total)<br><input id=\"dmt-cached-tokens\" type=\"number\" min=\"0\" step=\"1\" value=\"2000000\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Cache-write tokens (total)<br><input id=\"dmt-cache-write-tokens\" type=\"number\" min=\"0\" step=\"1\" value=\"0\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Output tokens (total)<br><input id=\"dmt-output-tokens\" type=\"number\" min=\"0\" step=\"1\" value=\"1000000\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Input $\/1M<br><input id=\"dmt-input-rate\" type=\"number\" min=\"0\" step=\"0.001\" value=\"0.2\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Cached input $\/1M<br><input id=\"dmt-cached-rate\" type=\"number\" min=\"0\" step=\"0.001\" value=\"0.02\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Cache write $\/1M<br><input id=\"dmt-cache-write-rate\" type=\"number\" min=\"0\" step=\"0.001\" value=\"0.25\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Output $\/1M<br><input id=\"dmt-output-rate\" type=\"number\" min=\"0\" step=\"0.001\" value=\"1.2\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Tool calls (total)<br><input id=\"dmt-tool-calls\" type=\"number\" min=\"0\" step=\"1\" value=\"150\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Tool price $\/1K calls<br><input id=\"dmt-tool-rate\" type=\"number\" min=\"0\" step=\"0.01\" value=\"2.5\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Other API\/infra cost $<br><input id=\"dmt-other-cost\" type=\"number\" min=\"0\" step=\"0.01\" value=\"5\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Reviewer hours (total)<br><input id=\"dmt-review-hours\" type=\"number\" min=\"0\" step=\"0.1\" value=\"3\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Reviewer cost $\/hour<br><input id=\"dmt-review-rate\" type=\"number\" min=\"0\" step=\"0.01\" value=\"25\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n    <label>Correction\/rework cost $<br><input id=\"dmt-rework-cost\" type=\"number\" min=\"0\" step=\"0.01\" value=\"20\" style=\"width:100%;background:#fff;color:#111827;border:1px solid #cbd5e1\"><\/label>\n  <\/div>\n  <button id=\"dmt-calculate\" type=\"button\" style=\"margin-top:16px;padding:10px 16px;border:0;border-radius:6px;background:#0b57d0;color:white;font-weight:700;cursor:pointer\">Calculate<\/button>\n  \n\n<p id=\"dmt-cost-error\" role=\"alert\" style=\"color:#a40000;font-weight:700\"><\/p>\n\n\n  <div id=\"dmt-cost-results\" aria-live=\"polite\" style=\"display:grid;grid-template-columns:repeat(auto-fit,minmax(190px,1fr));gap:12px;margin-top:12px\"><\/div>\n<\/div>\n<script>\n(function(){\n  var presets={\n    luna:[0.2,0.02,0.25,1.2],\n    terra:[2,0.2,2.5,12],\n    sol:[5,0.5,6.25,30]\n  };\n  function n(id){return Number(document.getElementById(id).value||0);}\n  function usd(v){return new Intl.NumberFormat('en-US',{style:'currency',currency:'USD',maximumFractionDigits:4}).format(v);}\n  function setPreset(){\n    var p=presets[document.getElementById('dmt-preset').value];\n    if(!p)return;\n    document.getElementById('dmt-input-rate').value=p[0];\n    document.getElementById('dmt-cached-rate').value=p[1];\n    document.getElementById('dmt-cache-write-rate').value=p[2];\n    document.getElementById('dmt-output-rate').value=p[3];\n  }\n  function calculate(){\n    var runs=n('dmt-runs'), accepted=n('dmt-accepted');\n    var error=document.getElementById('dmt-cost-error');\n    if(runs<1||accepted<0||accepted>runs){error.textContent='Accepted results must be between 0 and attempted runs.';return;}\n    error.textContent='';\n    var input=(n('dmt-input-tokens')\/1000000)*n('dmt-input-rate');\n    var cached=(n('dmt-cached-tokens')\/1000000)*n('dmt-cached-rate');\n    var cacheWrite=(n('dmt-cache-write-tokens')\/1000000)*n('dmt-cache-write-rate');\n    var output=(n('dmt-output-tokens')\/1000000)*n('dmt-output-rate');\n    var tools=(n('dmt-tool-calls')\/1000)*n('dmt-tool-rate');\n    var review=n('dmt-review-hours')*n('dmt-review-rate');\n    var machine=input+cached+cacheWrite+output+tools+n('dmt-other-cost');\n    var total=machine+review+n('dmt-rework-cost');\n    var acceptance=runs?accepted\/runs:0;\n    var cpa=accepted?total\/accepted:Infinity;\n    var items=[\n      ['Acceptance rate',(acceptance*100).toFixed(1)+'%'],\n      ['Model + tool + infra',usd(machine)],\n      ['Human review',usd(review)],\n      ['Total evaluation cost',usd(total)],\n      ['Cost per attempted run',usd(total\/runs)],\n      ['Cost per accepted result',accepted?usd(cpa):'No accepted result']\n    ];\n    document.getElementById('dmt-cost-results').innerHTML=items.map(function(x){return '<div style=\"background:white;color:#111827;border:1px solid #d7dce1;border-radius:8px;padding:12px\"><small>'+x[0]+'<\/small><br><strong style=\"font-size:1.25rem\">'+x[1]+'<\/strong><\/div>';}).join('');\n  }\n  document.getElementById('dmt-preset').addEventListener('change',function(){setPreset();calculate();});\n  document.getElementById('dmt-calculate').addEventListener('click',calculate);\n  calculate();\n})();\n<\/script>\n\n\n\n\n\n<p><strong>Important input rule:<\/strong> do not double-count cached tokens. Put regular\/billable non-cached input in \u201cInput tokens,\u201d cached reads in \u201cCached input tokens,\u201d and cache writes in their own field. Vendor usage exports differ; map each field to the current invoice documentation before using the output for finance.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">The formula<\/h2>\n\n\n\n\n\n\n\n<pre class=\"wp-block-code\"><code>Model cost =\n  (input tokens \/ 1,000,000 \u00d7 input rate)\n+ (cached input tokens \/ 1,000,000 \u00d7 cached-input rate)\n+ (cache-write tokens \/ 1,000,000 \u00d7 cache-write rate)\n+ (output tokens \/ 1,000,000 \u00d7 output rate)\n\nTool cost = tool calls \/ 1,000 \u00d7 tool-call rate\n\nHuman cost = reviewer hours \u00d7 loaded hourly cost\n\nTotal evaluation cost =\n  model cost + tool cost + other infrastructure\n+ human cost + correction\/rework cost\n\nAcceptance rate = accepted results \/ attempted runs\n\nCost per accepted result = total evaluation cost \/ accepted results<\/code><\/pre>\n\n\n\n<p>If no result is accepted, cost per accepted result is undefined\/infinite\u2014not zero. That outcome is informative: the workflow failed the rubric and should not be compared as if it produced usable work.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Current GPT-5.6 price presets<\/h2>\n\n\n\n\n\n\n\n<p>The calculator presets use OpenAI\u2019s standard short-context rates captured on August 11, 2026. Prices are USD per 1 million tokens.<\/p>\n\n\n\n\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Model<\/th><th>Input<\/th><th>Cached input<\/th><th>Cache write<\/th><th>Output<\/th><\/tr><\/thead><tbody><tr><td>GPT-5.6 Luna (short)<\/td><td>$0.20<\/td><td>$0.02<\/td><td>$0.25<\/td><td>$1.20<\/td><\/tr><tr><td>GPT-5.6 Terra (short)<\/td><td>$2.00<\/td><td>$0.20<\/td><td>$2.50<\/td><td>$12.00<\/td><\/tr><tr><td>GPT-5.6 Sol (short)<\/td><td>$5.00<\/td><td>$0.50<\/td><td>$6.25<\/td><td>$30.00<\/td><\/tr><tr><td>GPT-5.6 Luna (long)<\/td><td>$0.40<\/td><td>$0.04<\/td><td>$0.50<\/td><td>$1.80<\/td><\/tr><tr><td>GPT-5.6 Terra (long)<\/td><td>$4.00<\/td><td>$0.40<\/td><td>$5.00<\/td><td>$18.00<\/td><\/tr><tr><td>GPT-5.6 Sol (long)<\/td><td>$10.00<\/td><td>$1.00<\/td><td>$12.50<\/td><td>$45.00<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\n<p>Switch the calculator to Custom for long-context or future rates. OpenAI also documents a 10% uplift for eligible regional-processing configurations. Do not apply that uplift unless your deployment actually uses that option.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Tool and infrastructure costs to include<\/h2>\n\n\n\n\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Cost component<\/th><th>Current OpenAI reference<\/th><th>How to enter it<\/th><\/tr><\/thead><tbody><tr><td>Tool calls<\/td><td>$2.50 per 1,000 calls on the captured pricing table<\/td><td>Use tool calls and tool rate fields<\/td><\/tr><tr><td>Web search<\/td><td>$10 per 1,000 calls plus search-content tokens where applicable<\/td><td>Add call charge to other cost or use a weighted tool rate; keep search tokens in token totals<\/td><\/tr><tr><td>Containers<\/td><td>Starts at $0.03 for 1 GB per priced session period; size\/time rules apply<\/td><td>Add the billed container total from usage, not an estimate from wall-clock alone<\/td><\/tr><tr><td>File storage<\/td><td>$0.10\/GB\/day after the first free GB on the captured page<\/td><td>Add the billed storage amount<\/td><\/tr><tr><td>External SaaS\/API<\/td><td>Provider-specific<\/td><td>Add actual call, seat or data cost to other cost<\/td><\/tr><tr><td>Compute\/orchestration<\/td><td>Your platform-specific amount<\/td><td>Add cloud\/run cost to other cost<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\n<p>Built-in tool tokens are generally charged at the selected model\u2019s token rates even when a separate call fee also applies. Review the official page for the exact tool because billing units and minimums can change.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Define \u201caccepted\u201d before the test<\/h2>\n\n\n\n\n\n\n\n<p>An acceptance rate is meaningless if reviewers decide after seeing which model produced the output. Use the same rubric, blinded where practical, with objective thresholds.<\/p>\n\n\n\n\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Dimension<\/th><th>Pass rule example for a source-backed article package<\/th><\/tr><\/thead><tbody><tr><td>Factual support<\/td><td>Every material claim maps to a current primary source; no invented quote\/date\/number<\/td><\/tr><tr><td>Intent<\/td><td>Answers the assigned primary query and reader job without changing scope<\/td><\/tr><tr><td>Completeness<\/td><td>All required sections, examples, limitations and implementation details are present<\/td><\/tr><tr><td>Duplication<\/td><td>New information gain is material; existing owners are linked rather than rewritten<\/td><\/tr><tr><td>Rights<\/td><td>Every visual has source, capture date, provenance, alt text and credit<\/td><\/tr><tr><td>Technical validation<\/td><td>Code\/tool\/template tests pass in the stated environment<\/td><\/tr><tr><td>Editorial quality<\/td><td>No major rewrite required; voice and structure meet the publishing standard<\/td><\/tr><tr><td>Release safety<\/td><td>Metadata, canonical, internal links and rendered QA pass<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\n<p>Decide whether a correctable output is accepted. A useful policy is: accepted means no material factual\/intent error and no more than a defined correction threshold, such as ten reviewer minutes. Track correction time separately even when the item passes.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Evaluation data template<\/h2>\n\n\n\n\n\n\n\n<pre class=\"wp-block-code\"><code>run_id,model,reasoning,prompt_version,task_id,input_tokens,cached_input_tokens,cache_write_tokens,output_tokens,tool_calls,tool_cost,infra_cost,latency_seconds,retry_of,review_minutes,correction_minutes,accepted,rejection_reason\n001,gpt-5.6-luna,high,v3,brief-01,48000,12000,0,9600,2,0.005,0.03,82,,3,0,true,\n002,gpt-5.6-luna,high,v3,brief-02,51000,14000,0,10200,3,0.0075,0.03,94,,11,8,false,unsupported_claim\n003,gpt-5.6-terra,high,v3,brief-02,56000,15000,0,11400,3,0.0075,0.03,106,002,4,0,true,<\/code><\/pre>\n\n\n\n<p>Keep task IDs and prompt versions so you can separate model effects from a changing test. A retry belongs to the same reader job and its cost remains in the numerator.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Worked example<\/h2>\n\n\n\n\n\n\n\n<p>The calculator defaults to an illustrative 100-run Luna batch: 5 million regular input tokens, 2 million cached input tokens, 1 million output tokens, 150 generic tool calls, $5 other infrastructure, three reviewer hours at $25\/hour and $20 rework. If 80 results pass, the model\/token portion is small relative to human review. That is common in low-priced models: optimizing a few cached-token cents matters less than reducing correction time or increasing acceptance.<\/p>\n\n\n\n\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Optimization<\/th><th>Can reduce cost<\/th><th>Can increase cost\/risk<\/th><\/tr><\/thead><tbody><tr><td>Use a cheaper worker model<\/td><td>Lower token bill<\/td><td>More retries or reviewer correction<\/td><\/tr><tr><td>Cache stable context<\/td><td>Lower repeated input cost<\/td><td>Stale context if invalidation is weak<\/td><\/tr><tr><td>Increase reasoning<\/td><td>Higher first-pass acceptance on difficult tasks<\/td><td>More tokens\/latency on simple tasks<\/td><\/tr><tr><td>Add deterministic validation<\/td><td>Catches failures before human review<\/td><td>Engineering and maintenance cost<\/td><\/tr><tr><td>Reduce source scope<\/td><td>Less token\/tool work<\/td><td>Missed evidence if scope becomes too narrow<\/td><\/tr><tr><td>Use stronger model only on escalations<\/td><td>Balances worker cost and difficult cases<\/td><td>Routing errors if escalation rules are vague<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Compare configurations fairly<\/h2>\n\n\n\n\n\n\n\n<ol class=\"wp-block-list\"><li>Freeze a representative task set before testing.<\/li><li>Use the same acceptance rubric and reviewer instructions.<\/li><li>Record model, reasoning, prompt version, tools and context mode.<\/li><li>Include failed runs and retries.<\/li><li>Measure human time, not just API spend.<\/li><li>Report confidence intervals or at least the sample size; ten easy tasks are not a production benchmark.<\/li><li>Segment by task difficulty instead of averaging unlike work.<\/li><li>Run sensitivity scenarios for reviewer rate, acceptance and token prices.<\/li><\/ol>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Sensitivity analysis: what usually moves the answer<\/h2>\n\n\n\n\n\n\n\n<p>Do not present one calculator output as a forecast. Change the assumptions that are both uncertain and economically important. In many marketing-agent workflows, reviewer time and acceptance rate dominate the token rate.<\/p>\n\n\n\n\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Scenario<\/th><th>Change<\/th><th>Question answered<\/th><\/tr><\/thead><tbody><tr><td>Base<\/td><td>Observed token\/tool\/review totals<\/td><td>What did the test actually cost?<\/td><\/tr><tr><td>Acceptance downside<\/td><td>Acceptance rate minus 15 points<\/td><td>How expensive is quality drift?<\/td><\/tr><tr><td>Review downside<\/td><td>Reviewer minutes +50%<\/td><td>What if corrections are understated?<\/td><\/tr><tr><td>Scale<\/td><td>10\u00d7 tasks, separate fixed and variable cost<\/td><td>Does orchestration amortize or do review bottlenecks grow?<\/td><\/tr><tr><td>Model escalation<\/td><td>Worker model for 80%; stronger model for 20%<\/td><td>Is routing better than one model for everything?<\/td><\/tr><tr><td>Price change<\/td><td>Current vendor rates \u00b120%<\/td><td>Would the decision survive a pricing update?<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\n<p>Cost per accepted result is nonlinear with acceptance: the denominator shrinks as quality falls while failed-run cost remains. That is why a model with one-tenth the token price is not automatically ten times cheaper per useful output.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Spreadsheet formulas<\/h2>\n\n\n\n\n\n\n\n<p>If your columns are attempted runs in <code>B2<\/code>, accepted results in <code>C2<\/code>, total model\/tool\/infra cost in <code>D2<\/code>, reviewer hours in <code>E2<\/code>, hourly rate in <code>F2<\/code> and rework cost in <code>G2<\/code>:<\/p>\n\n\n\n\n\n\n\n<pre class=\"wp-block-code\"><code>Acceptance rate: =IFERROR(C2\/B2,0)\nHuman review cost: =E2*F2\nTotal evaluation cost: =D2+(E2*F2)+G2\nCost per attempt: =IFERROR((D2+(E2*F2)+G2)\/B2,\"\")\nCost per accepted result: =IF(C2=0,\"No accepted result\",(D2+(E2*F2)+G2)\/C2)<\/code><\/pre>\n\n\n\n<p>Keep raw runs on one sheet and aggregate by model, prompt version and task segment with a pivot table. Never average the \u201ccost per accepted result\u201d ratios from unequal batches; sum costs and accepted counts first, then divide.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Decision thresholds<\/h2>\n\n\n\n\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Ship:<\/strong> acceptance and incident thresholds pass, cost is within the business ceiling and the sample represents production.<\/li><li><strong>Route:<\/strong> cheap worker passes routine tasks but difficult segments need escalation.<\/li><li><strong>Redesign:<\/strong> reviewer time or repeated failures dominate; change the task, data or validation before changing the model.<\/li><li><strong>Stop:<\/strong> no accepted outputs, unsafe incidents, unmeasurable rubric or economics worse than the existing manual process.<\/li><\/ul>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Metrics to report beside cost<\/h2>\n\n\n\n\n\n\n\n<ul class=\"wp-block-list\"><li>acceptance rate and rejection reasons;<\/li><li>cost per attempt and per accepted result;<\/li><li>median and tail latency;<\/li><li>review\/correction minutes per accepted result;<\/li><li>retry and escalation rate;<\/li><li>tool-denial or policy-failure rate;<\/li><li>quality score by task segment;<\/li><li>material incident count.<\/li><\/ul>\n\n\n\n\n\n\n\n<p>A configuration with a slightly higher cost but much lower incident rate can be the correct business choice. The calculator is an economic lens, not a substitute for safety or quality gates.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">FAQ<\/h2>\n\n\n\n\n\n\n\n<h3 class=\"wp-block-heading\">Should retries be included?<\/h3>\n\n\n\n\n\n\n<p>Yes. A retry is part of the cost of obtaining the accepted result. Excluding failed attempts systematically favours unreliable workflows.<\/p>\n\n\n\n\n\n\n\n<h3 class=\"wp-block-heading\">Should reviewer time be included?<\/h3>\n\n\n\n\n\n\n<p>Yes. Use a loaded hourly cost or a consistent internal rate. Reviewer time often exceeds token cost and can reverse the apparent model ranking.<\/p>\n\n\n\n\n\n\n\n<h3 class=\"wp-block-heading\">Is cost per accepted result the same as ROI?<\/h3>\n\n\n\n\n\n\n<p>No. Cost per accepted result measures production efficiency. ROI also needs the value or incremental benefit of the accepted result, attribution, operating period and counterfactual.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">How this calculator was verified<\/h2>\n\n\n\n\n\n\n<p>DMT transcribed the GPT-5.6 and tool rates from OpenAI\u2019s official pricing page captured August 11, 2026, kept every rate editable, and checked the JavaScript formula against the displayed equations and worked inputs. The calculator does not fetch billing data and cannot know discounts, contract rates or later price changes.<\/p>\n\n\n\n\n\n\n\n<h2 class=\"wp-block-heading\">Bottom line<\/h2>\n\n\n\n\n\n\n\n<p>Measure the whole system: tokens, cache, tools, infrastructure, retries, review and correction. Define acceptance first and count every attempt. Then use the calculator to compare configurations on the same task set. For workflow design and cost controls around Luna, continue with DMT\u2019s <a href=\"https:\/\/dmarketertayeeb.com\/blog\/gpt-5-6-luna-price-cut-ai-marketing-automation\/\">Luna automation cost playbook<\/a>.<\/p>\n\n\n\n\n\n\n\n<p>Use the result alongside the <a href=\"https:\/\/dmarketertayeeb.com\/blog\/openai-gpt-5-6-luna-terra-price-cuts\/\">GPT-5.6 pricing-change analysis<\/a>, <a href=\"https:\/\/dmarketertayeeb.com\/blog\/gpt-5-6-sol-terra-luna-marketers-guide\/\">model-routing guide<\/a>, <a href=\"https:\/\/dmarketertayeeb.com\/blog\/luna-max-codex-subagents-sol-high\/\">worker\/reviewer pattern<\/a> and <a href=\"https:\/\/dmarketertayeeb.com\/blog\/openai-codex-marketers-plugins-sites-workflows\/\">Work\/Codex automation architecture<\/a>.<\/p>\n\n\n","protected":false},"excerpt":{"rendered":"<p>A working AI agent cost-per-accepted-result calculator with current GPT-5.6 price presets, tool costs, reviewer time, acceptance rate and an evaluation template.<\/p>\n","protected":false},"author":1,"featured_media":2777,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[386],"tags":[411,412,414,413,195],"class_list":["post-2778","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-for-marketers","tag-ai-agent-cost","tag-ai-calculator","tag-ai-evaluation","tag-gpt-5-6-pricing","tag-marketing-automation","has-featured-image"],"_links":{"self":[{"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/posts\/2778","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/comments?post=2778"}],"version-history":[{"count":2,"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/posts\/2778\/revisions"}],"predecessor-version":[{"id":2788,"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/posts\/2778\/revisions\/2788"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/media\/2777"}],"wp:attachment":[{"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/media?parent=2778"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/categories?post=2778"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dmarketertayeeb.com\/blog\/wp-json\/wp\/v2\/tags?post=2778"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}