Se connecter
Claude Sonnet 5 logo

Claude Sonnet 5

Anthropic · Sorti juin 2026

Voir les tarifsComparer les modèles
Recommandé

Anthropic's most agentic Sonnet at launch delivers performance close to Opus 4.8 at 40 percent of the Opus rate. The $2/$10 per 1M token price is the standard rate: Anthropic cancelled the increase to $3/$15 that was scheduled for 1 September 2026. The newer tokenizer produces about 30 percent more tokens for the same text, so recount budgets before migrating from Sonnet 4.6. Anthropic now lists Sonnet 5 as legacy; Claude Sonnet 5.5, at the same $2/$10, is the current Sonnet, so start new work there.

Est-ce fait pour vous ?

Recommandé pour

  • Multi-step agentic coding and autonomous task completion — finishes complex workflows previous Sonnet models stalled on
  • Brownfield debugging: traces failures to root causes, writes reproducing tests, and ships durable fixes unprompted
  • Production AI agents at attractive cost — near-Opus 4.8 quality on agentic benchmarks at 40% of the price ($2/$10 against Opus 4.8's $5/$25)
  • Knowledge work where it edges Opus 4.8 (GDPval-AA v2: 1,618 vs 1,615) — document drafting, research, professional analysis
  • Self-verification: checks its own output without being explicitly prompted, reducing compounding errors in automated pipelines
  • High-volume batch pipelines: the Batch API halves the rate to $1 input / $5 output per 1M, and cache reads land at $0.20 per 1M

Déconseillé pour

  • Maximum-capability frontier tasks where Opus 5, Fable 5 or Fable 5.1 are necessary — Sonnet 5 gets close but doesn't match the top tier
  • Cybersecurity work requiring reduced guardrails — cyber safeguards are enabled by default and Anthropic recommends the Opus tier for that use case
  • Workloads where the new tokenizer's ~30% token inflation matters — recount budgets before switching; same text costs more tokens than on Sonnet 4.6

Performances par tâche

Codage agentique

Excellent

Terminal-Bench 2.1 80,4% (+13,4 pts vs Sonnet 4.6) — le signal le plus clair de fiabilité agentique multi-étapes. Benchmark production Cursor : 61,2% (vs 49,0% pour 4.6).

Génération de code

Very Good

SWE-bench Verified 85,2% — 4,1 pts d'avance sur Sonnet 4.6 (81,1%), réduisant l'écart avec Opus 4.8 (88,6%). Solide en codage standard mais Opus garde l'avance sur Pro de 6 pts.

Débogage

Excellent

Les premiers testeurs soulignent régulièrement les tests de reproduction spontanés, le stash pour confirmer les régressions et le traçage des causes racines sur du code existant.

Utilisation d'outils et d'ordinateur

Excellent

OSWorld-Verified 81,2%, équivalent à Opus 4.8 (83,4%) et nettement devant Sonnet 4.6 (78,5%). Gère les navigateurs, terminaux et outils API de bout en bout.

Travail de connaissance

Excellent

GDPval-AA v2 : 1 618 — devance légèrement Opus 4.8 (1 615). Une véritable première pour un modèle Sonnet de milieu de gamme face à son grand frère flagship.

Raisonnement

Very Good

Humanity's Last Exam : 43,2% (sans outils) / 57,4% (avec outils). Avec outils, se rapproche d'Opus 4.8 (57,9%) ; sans outils, reste derrière le niveau flagship.

Cybersécurité

Fair

Par conception — n'a jamais développé d'exploit fonctionnel complet sur les tests de vulnérabilité Firefox 147 (0,0%). Garde-fous cyber activés par défaut. Anthropic recommande Opus 4.8 pour le travail cyber.

Tarifs

Entrée

$2 / 1M

Sortie

$10 / 1M

Contexte

1M tokens; batch $1/$5; cache read $0.20

Voir tous les tarifs

Tests de performance

TestScoreSource
SWE-bench Verified85.2% Source
Terminal-Bench 2.180.4% Source
SWE-bench Pro63.2% Source
Humanity's Last Exam (no tools)43.2% Source
Humanity's Last Exam (with tools)57.4% Source
OSWorld-Verified81.2% Source
Knowledge Work (GDPval-AA v2)1,618 Source
Cursor Production Benchmark (Max effort)61.2% Source
BrowseComp 25 (agentic search)84.7% Source

Historique des verdicts

30 sept. 2026
recommended -> recommended (summary corrected: legacy status, unsourced plan-default claim removed): Rating unchanged. The standing summary said Sonnet 5 is the default model on Free and Pro plans; Anthropic's pricing page does not say so and lists Sonnet 5 as legacy behind Sonnet 5.5 at the same $2/$10 rate. The claim is removed and the legacy status stated.
10 août 2026
recommended -> recommended (summary corrected: $2/$10 is the standard price, 40 percent of the Opus rate): Rating unchanged. The standing summary described $2/$10 as introductory pricing expiring 31 August 2026 and put Sonnet 5 at roughly half the Opus price. Anthropic's live rate card confirms $2/$10 is now the standard price and the $3/$15 increase will not occur, which puts Sonnet 5 at 40 percent of the $5/$25 Opus rate.
15 juil. 2026
copy edit to brevity standard, detail preserved in children: EN verdict_summary was ~116 words / 5 sentences — over the 2–4 sentence / ≤90 word target. Shortened to 3 sentences / ~70 words while preserving all claims. Children (strengths, task_strengths, etc.) retain the full detail.

Sources

Journal de vérification

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Contenu· Mis à jour

    Agent automatisé

    Rate $2/$10 unchanged. Sonnet 5 is now legacy behind Sonnet 5.5 at the same rate. Unsourced 'default model on Free and Pro plans' claim removed from summary (verdict_set, recommended held) and audiences; legacy note added.

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

    Sonnet 5 $2 in / $10 out, cache hits $0.20, batch $1/$5. Standard-price note verbatim: the $2/$10 pricing 'is now the standard price. The previously scheduled increase to $3/$15 ... on September 1, 2026 will not occur.' Matches stored. No update. Source row relabelled from the misleading 'Claude Platform on AWS pricing docs' (7ef6c662).

  • Tarifs· Aucun changement

    Agent automatisé

    Rate card: Sonnet 5 $2 in / $10 out, cache hits $0.20, batch $1/$5; standard-price note unchanged. Matches stored scalars and pricing_context. Trigger CLEARED: the 09-09 summary payload landed at WF-prod 09-14 (verdict_history 'summary corrected: $2/$10 is the standard price'); live summary no longer calls $2/$10 intro pricing. No update.

  • Tarifs· Aucun changement

    Agent automatisé

    Rate card: Claude Sonnet 5 $2/MTok input, $10/MTok output, cache hits $0.20, batch $1/$5. Note verbatim: 'The $2/$10 ... announced at launch as introductory pricing through August 31, 2026, is now the standard price. The previously scheduled increase to $3/$15 ... on September 1, 2026 will not occur.' Matches stored $2/$10 and pricing_context. The dispatch's '$3/$15 directory vs $2/$10 live' drift item is stale: scalars were fixed 2026-09-09. No update. The summary still says 'intro pricing ($2/$10 per MTok through Aug 31)'; its 2026-09-09 payload on draft claude-sonnet-5-price-cut-made-permanent has NOT landed (verdict_history latest 2026-07-15). Not re-staged. Cross-entity claude-4-sonnet 'same rate card' prose not checked this run.

  • Tarifs· Mis à jour

    Agent automatisé

    LIVE WRONG PRICE CORRECTED on a high-traffic entity. Stored pricing_input/$pricing_output read $3 / 1M and $15 / 1M; Anthropic's live rate card reads $2 / MTok input and $10 / MTok output. The stored figure was never a live rate: it was the announced 1 September 2026 successor to the launch price, written in ahead of a change Anthropic has since cancelled. The page carries an explicit note: 'The $2/$10 ... announced at launch as introductory pricing through August 31, 2026, is now the standard price. The previously scheduled increase to $3/$15 ... on September 1, 2026 will not occur.' Corrected pricing_input to $2 / 1M, pricing_output to $10 / 1M, and set the per_mtok decimals to 2.00 / 10.00. Refreshed pricing_context to '1M tokens; batch $1/$5; cache read $0.20' (batch $1/$5 and cache read $0.20 both read off the same table). Rewrote the strength that claimed 'roughly half the price' of Opus 4.8 to the correct 40% ($2/$10 against $5/$25), added a batch-pipeline best_for, and replaced the two audience rows that framed $2/$10 as intro pricing expiring 31 August. Rating stays recommended. Verdict summary still carries the stale intro-pricing framing and the cancelled rate; that rewrite is unreachable from model_update and is staged as an entity_verdict payload on draft post 'claude-sonnet-5-price-cut-made-permanent'.

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

    Live $2/$10 intro through Aug 31; stored $3/$15 standard rate. No change to standard rate.

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Importé au lancement

  • Profil· Aucun changement

    Importé au lancement

Comment nous évaluons