Gemini 4 logo

Gemini 4

Google · Sorti sept. 2026

pending

Le prochain modèle de Google DeepMind, confirmé. Le pré-entraînement a débuté le 21 juillet 2026. Aucun benchmark, aucun tarif, aucune spécification, aucune date de GA.

Est-ce fait pour vous ?

Recommandé pour

  • Agentic coding, enterprise knowledge work and cyber defense, the stated design focus of Gemini 4 Argon, per Google's own self-reported benchmarks
  • Trying a frontier model at the introductory $2 input and $10 output per 1M tokens, which Google's pricing page lists through Nov 30 before the rate moves to $4 and $20

Déconseillé pour

  • Budgets that assume the introductory rate past Nov 30: Google's pricing page lists $4 input and $20 output per 1M tokens after that date
  • Decisions that need independent benchmark confirmation: every published score is self-reported by Google

Tarifs

Entrée

$2 / 1M tokens (through Nov 30)

Sortie

$10 / 1M tokens (through Nov 30)

Contexte

$2/$10 to Nov 30, then $4/$20; paid tier only

Voir tous les tarifs

Tests de performance

TestScoreSource
DeepSWE v1.177.9% Source
CWE-bench v168.0% (tied first) Source
LVBench91.7% Source
AutomationBench51.3% (#1) Source
Vals Index68.9% Source

Aucun changement de verdict pour l’instant

L’horloge tourne dès le premier jour : les changements apparaîtront ici à mesure que notre verdict évolue.

Journal de vérification

  • Tarifs· Mis à jour

    Agent automatisé

    Gemini 4 Argon is generally available since Oct 13, 2026 per the Gemini API changelog. Pricing $2/$10 per 1M through Nov 30, then $4/$20. Refreshed pricing fields, context window, strengths, audiences and sources. Standing verdict_summary still says pre-GA with no pricing; staged pending->conditional at publish.

  • Contenu· Mis à jour

    Agent automatisé

    Google announced Gemini 4 Argon on 2026-09-30: limited rollout to trusted cyber defenders (Fairwind Program), paid API and AI Ultra next, no GA date. Self-reported benchmarks added. Release date, pricing, strengths, audiences and sources refreshed. Stored verdict_summary still describes pre-training and is staged for the publish step.

  • Tarifs· Mis à jour

    Agent automatisé

    Pricing went from Not yet published to introductory $2 input and $10 output per 1M tokens, cached input 95% off input price.

Comment nous évaluons