Se connecter
Claude Sonnet 4.6 logo

Claude Sonnet 4.6

Anthropic · Sorti févr. 2026

Voir les tarifsComparer les modèles
Conditionnel

Superseded by Claude Sonnet 5 (shipped Jun 30, 2026), which Anthropic now lists as legacy behind Claude Sonnet 5.5 as the current Sonnet. Both list at $2/$10 per 1M tokens against Sonnet 4.6's $3/$15, since Anthropic cancelled the scheduled rise to $3/$15. Sonnet 5 also adds a 13.4-point Terminal-Bench 2.1 jump. Choose 4.6 only for pinned deployments or where Sonnet 5's ~30% higher token count and removed temperature/top_p/top_k are disqualifying.

Est-ce fait pour vous ?

Recommandé pour

  • Pinned or reproducibility-sensitive deployments: teams that need a stable, unchanged model behind their pipeline rather than migrating to Sonnet 5
  • Balanced production coding and document analysis: strong speed/quality/cost tradeoff at $3/$15 with a 1M-token context window
  • Code that depends on temperature/top_p/top_k, which Sonnet 5 removed

Déconseillé pour

  • New default workloads: Claude Sonnet 5.5 is Anthropic's current Sonnet at $2/$10 per 1M against Sonnet 4.6's $3/$15; start new work there (Sonnet 5, also $2/$10, is now legacy)
  • Longest-horizon agentic coding: Sonnet 5 finishes multi-step tasks 4.6 stalled on; Opus 4.8 / Fable 5 lead the frontier tier

Performances par tâche

Génération de code

Very Good

Solide pour sa génération ; Sonnet 5 termine désormais les tâches multi-étapes sur lesquelles 4.6 bloquait

Analyse de documents

Very Good

Le contexte de 1M tokens gère bien les grands documents ; bon équilibre entre qualité et coût

Codage agentique

Good

Capable mais supplanté — le bond de 13,4 points de Sonnet 5 sur Terminal-Bench 2.1 est la raison de migrer

Tarifs

Entrée

$3 / 1M tokens

Sortie

$15 / 1M tokens

Contexte

1M tokens

Voir tous les tarifs

Tests de performance

TestScoreSource
SWE-bench Verified72.7% Source
MMLU-Pro84.2% Source
HumanEval93.7% Source

Historique des verdicts

30 sept. 2026
conditional -> conditional (summary corrected: Sonnet 5.5 is now the current Sonnet): Rating unchanged. The standing summary called Sonnet 5 the default mid-tier workhorse. Anthropic's models overview lists Sonnet 5 as legacy and Sonnet 5.5 as current, both at $2/$10.
18 sept. 2026
conditional -> conditional (summary corrected: Sonnet 5 costs less, not the same): Rating unchanged. The standing summary said Sonnet 5 matched Sonnet 4.6 'at the same rate card'. Anthropic's rate card, read 2026-09-18, lists Sonnet 5 at $2/$10 and Sonnet 4.6 at $3/$15, with the note that the previously scheduled increase to $3/$15 'will not occur'.
15 juil. 2026
copy edit to brevity standard: detail preserved in children

Sources

Journal de vérification

  • Contenu· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Statut· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Contenu· Mis à jour

    Agent automatisé

    Rate $3/$15 unchanged. Summary and audiences called Sonnet 5 the default; Anthropic lists Sonnet 5 as legacy and Sonnet 5.5 as current at $2/$10. Summary corrected via verdict_set (conditional held); strengths, audiences and sources refreshed.

  • Tarifs· Aucun changement

    Agent automatisé

  • Profil· Mis à jour

    Agent automatisé

    Cross-entity coupling from the claude-sonnet-5 trigger. Rate card read 2026-09-18: Claude Sonnet 4.6 $3/MTok in, $15/MTok out (matches stored scalars); Claude Sonnet 5 $2/$10, note verbatim 'The previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur.' The entity said Sonnet 5 was 'at the same rate card' in the summary, an avoid_for strength and a not_for_whom audience. That is false: Sonnet 5 is cheaper. FIXED via model_update: the two rate-card rows now say $2/$10 against $3/$15; removed the for_whom row claiming the ~30% tokenizer expansion would raise effective cost (1.3x $2/$10 is still below $3/$15, so that claim was unsupported) and replaced it with a neutral measure-before-switching row; split the temperature/top_p row out of the cost row; pricing_url anthropic.com/pricing (unread) -> platform rate card (read). Summary corrected via verdict_set (same rating conditional, id 286348bf), no announcing post: a correction of our own prose, not vendor news. Scalars unchanged.

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Profil· Mis à jour

    Agent automatisé

    Rewrote seed-junk entry in place (slug kept): name 'Claude 4 Sonnet' -> 'Claude Sonnet 4.6'; release 2026-03-15 -> real 2026-02-17; context set to 1M; pricing confirmed $3/$15; verdict recommended -> conditional (superseded by Sonnet 5, still available for pinned/reproducible use). Summary/strengths/audiences/task-ratings/sources fully replaced with sourced Sonnet 4.6 predecessor data.

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

    Pricing unchanged: $3/$15 per 1M tokens. Current Sonnet tier is 4.6 on the pricing page.

  • Tarifs· Aucun changement

    Agent automatisé

    Sonnet 4 pricing unchanged at $3/$15 in legacy section. Current Sonnet 4.6 also at $3/$15.

  • Tarifs· Aucun changement

    Agent automatisé

  • Profil· Aucun changement

    Importé au lancement

  • Tarifs· Aucun changement

    Importé au lancement

Comment nous évaluons