Anthropic Claude Sonnet 5 vs Sonnet 4.6 vs Opus 4.8: Agentic Coding Benchmarks, API Pricing, and Cost-Performance Tradeoffs Compared

Anthropic just shipped Claude Sonnet 5. They call it its most agentic Sonnet model yet. It plans, drives browsers and terminals, and runs autonomously across long tasks. Sonnet 5 is the default model for Free and Pro plans today. Max, Team, and Enterprise users can select it. It is also live in Claude Code and on the Claude Platform. TL;DR Sonnet 5 is Anthropic’s most agentic mid-tier model, closing much of the gap to Opus 4.8. Beats Sonnet 4.6 on every published benchmark: 63.2% SWE-bench Pro, 81.2% OSWorld-Verified, 57.4% HLE. Cheaper to run: $2/$10 per MTok intro pricing through Aug 31, then $3/$15; Opus 4.8 is $5/$25. Best value at low/medium effort; at xhigh it can cost more than Opus 4.8 for similar quality. Safer than 4.6, with deliberately low cyber capability — Opus stays the pick for accuracy-critical work. Claude Sonnet 5 Sonnet sits in the middle of Anthropic’s lineup. It is above the cheaper Haiku 4.5 and below the flagship Opus 4.8. Sonnet 5 is an upgrade to Sonnet 4.6, which launched in February 2026. Anthropic frames this release around agentic reliability, not one headline benchmark. In practice, that means longer task chains without losing context. It means better self-correction when a tool call fails. It means steadier behavior across extended sessions inside Claude Code or Cowork. The model exposes effort levels: low, medium, high, and xhigh (extra high). Higher effort spends more tokens on reasoning. That raises both quality and cost. It is important to note that Sonnet 5 uses an updated tokenizer, the same one introduced with Opus 4.7. The same text can map to roughly 1.0 to 1.35 times more tokens. Interactive Explainer Claude Sonnet 5 Cost & Capability Explorer :root{ --bg:#ffffff; --panel:#f6f8f5; --panel2:#eef2ea; --ink:#16201a; --muted:#5b6b60; --line:#e0e6dd; --green:#5a8f1e; --greenDark:#3f6b12; --accent:#76B900; --s46:#9aa6ad; --s5:#76B900; --opus:#e0a800; } *{box-sizing:border-box} html,body{margin:0} #mtp-s5{ font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,Helvetica,Arial,sans-serif; color:var(--ink); background:var(--bg); padding:22px; max-width:880px; margin:0 auto; line-height:1.5; } #mtp-s5 h2{font-size:21px; margin:0 0 4px; letter-spacing:-.2px} #mtp-s5 .sub{color:var(--muted); font-size:13.5px; margin:0 0 18px} #mtp-s5 .card{ background:var(--panel); border:1px solid var(--line); border-radius:14px; padding:18px; margin-bottom:18px; } #mtp-s5 .card h3{font-size:15px; margin:0 0 14px; display:flex; align-items:center; gap:8px} #mtp-s5 .dot{width:9px;height:9px;border-radius:50%;background:var(--accent);display:inline-block} #mtp-s5 .row{display:flex; flex-wrap:wrap; gap:8px; margin-bottom:14px} #mtp-s5 .seg{display:flex; flex-wrap:wrap; gap:6px} #mtp-s5 button.opt{ font:inherit; font-size:12.5px; cursor:pointer; padding:7px 12px; border-radius:999px; border:1px solid var(--line); background:#fff; color:var(--ink); transition:.15s; } #mtp-s5 button.opt:hover{border-color:var(--accent)} #mtp-s5 button.opt[aria-pressed="true"]{background:var(--accent); border-color:var(--accent); color:#16201a; font-weight:600} #mtp-s5 label.lbl{display:block; font-size:12px; color:var(--muted); margin:0 0 6px} #mtp-s5 .grid2{display:grid; grid-template-columns:1fr 1fr; gap:16px} #mtp-s5 input[type=range]{width:100%; accent-color:var(--accent)} #mtp-s5 .val{font-variant-numeric:tabular-nums; font-weight:600} #mtp-s5 .costbox{display:flex; align-items:baseline; gap:10px; flex-wrap:wrap; margin-top:6px} #mtp-s5 .cost{font-size:30px; font-weight:700; color:var(--greenDark); font-variant-numeric:tabular-nums} #mtp-s5 .costnote{font-size:12.5px; color:var(--muted)} #mtp-s5 .pill{font-size:11.5px; background:var(--panel2); border:1px solid var(--line); border-radius:999px; padding:3px 9px; color:var(--muted)} #mtp-s5 .compare{display:flex; gap:10px; margin-top:12px; flex-wrap:wrap} #mtp-s5 .ctag{font-size:12px; color:var(--muted)} #mtp-s5 .ctag b{color:var(--ink)} #mtp-s5 .bar-wrap{margin:10px 0} #mtp-s5 .bar-row{display:flex; align-items:center; gap:10px; margin:9px 0} #mtp-s5 .bar-name{width:88px; font-size:12.5px; color:var(--muted); flex:none} #mtp-s5 .track{flex:1; background:#fff; border:1px solid var(--line); border-radius:8px; height:24px; position:relative; overflow:hidden} #mtp-s5 .fill{height:100%; border-radius:7px 0 0 7px; transition:width .5s ease; display:flex; align-items:center; justify-content:flex-end; padding-right:7px; font-size:11.5px; font-weight:700; color:#16201a} #mtp-s5 .legend{display:flex; gap:14px; flex-wrap:wrap; font-size:12px; color:var(--muted); margin-top:8px} #mtp-s5 .lg{display:inline-flex; align-items:center; gap:6px} #mtp-s5 .sw{width:12px;height:12px;border-radius:3px;display:inline-block} #mtp-s5 .callout{font-size:12.5px; color:var(--muted); background:var(--panel2); border-left:3px solid var(--accent); padding:9px 12px; border-radius:0 8px 8px 0; margin-top:12px} #mtp-s5 .foot{text-align:center; font-size:12px; color:var(--muted); margin-top:6px} #mtp-s5 .foot b{color:var(--green)} #mtp-s5 .metricbtns{margin-bottom:6px} @media (max-width:640px){ #mtp-s5{padding:16px} #mtp-s5 .grid2{grid-template-columns:1fr} #mtp-s5 .bar-name{width:74px} #mtp-s5 .cost{font-size:25px} } Claude Sonnet 5 — Cost & Capability Explorer Estimate per-task cost across models and compare published benchmarks. All figures from Anthropic’s June 30, 2026 launch. Per-task cost estimator Sonnet 5 (intro $2/$10) Sonnet 5 (std $3/$15) Opus 4.8 ($5/$25) Sonnet 4.6 ($3/$15) Input tokens per task: 20,000 Output tokens per task: 6,000 Tasks per day: 500 Sonnet 5 tokenizer factor: 1.15× $0.00 per task • $0.00/day • $0.00/mo Sonnet 5 uses an updated tokenizer (same as Opus 4.7). The same text can map to roughly 1.0–1.35× more tokens, so the factor is applied to Sonnet 5 only. Published benchmark comparison Agentic coding (SWE-bench Pro) Terminal-Bench 2.1 Computer use (OSWorld-Verified) Humanity’s Last Exam (tools) Sonnet 4.6 Sonnet 5 Opus 4.8 On knowledge work (GDPval-AA v2), Sonnet 5 scores 1,618 and edges Opus 4.8’s 1,615. That benchmark uses a different scale, so it is shown here as a note rather than a bar. Interactive explainer by Marktechpost • figures: Anthropic launch & system card, June 30, 2026 (function(){ var price={ s5i:{i:2,o:10}, s5s:{i:3,o:15}, opus:{i:5,o:25}, s46:{i:3,o:15} }; var names={ s5i:"Sonnet 5 intro", s5s:"Sonnet 5 std", opus:"Opus 4.8", s46:"Sonnet 4.6" }; var state={ model:"s5i" }; var bench={ swe:{ s46:58.1, s5:63.2, opus:69.2 }, term:{ s46:67.0, s5:80.4, opus:null }, osw:{ s46:78.5, s5:81.2, opus:null }, hle:{ s46:46.8, s5:57.4, opus:57.9 } }; var benchMetric="swe"; function $(id){return document.getElementById(id);} function fmt(n){return n.toLocaleString(undefined,{maximumFractionDigits:0});} function money(n){ if(n<0.01) return "$"+n.toFixed(4); if(n<1) return "$"+n.toFixed(3); return "$"+n.toLocaleString(undefined,{minimumFractionDigits:2,maximumFractionDigits:2}); } function tkFactor(){ return parseInt($("tk").value,10)/100; } function costFor(modelKey){ var inTok=parseInt($("inTok").value,10); var outTok=parseInt($("outTok").value,10); var f=1; if(modelKey==="s5i"||modelKey==="s5s"){ f=tkFactor(); } var p=price[modelKey]; return (inTok*f/1e6)*p.i + (outTok*f/1e6)*p.o; } function render(){ var tasks=parseInt($("tasks").value,10); $("inVal").textContent=fmt(parseInt($("inTok").value,10)); $("outVal").textContent=fmt(parseInt($("outTok").value,10)); $("taskVal").textContent=fmt(tasks); $("tkVal").innerHTML=tkFactor().toFixed(2)+"&times;"; var c=costFor(state.model); $("costOut").textContent=money(c); $("dayOut").textContent=money(c*tasks); $("moOut").textContent=money(c*tasks*30); // comparison vs other models at same workload var order=["s5i","s5s","opus","s46"]; var html=""; order.forEach(function(k){ if(k===state.model) return; var oc=costFor(k); var diff=oc===0?0:((c-oc)/oc*100); var sign=diff>0?"+":""; html+='<span class="ctag">vs <b>'+names[k]+'</b>: '+money(oc)+' ('+sign+diff.to