Se connecter
GLM-5.3 logo

GLM-5.3

Z.ai · Sorti août 2026

Voir les tarifsComparer les modèles
Conditionnel

GLM-5.3 is Z.ai's open-weights coding and cyber-defense model on the GLM-5.2 base, with every gain from post-training. Pricing is published at $1.40 input / $4.40 output per 1M, the same as GLM-5.2, and the weights are now on Hugging Face. It stays conditional because its headline coding and cyber scores come from Z.ai's own evaluations: reach for defensive security and agentic automation, but audit it yourself before standardizing.

Est-ce fait pour vous ?

Recommandé pour

  • Defensive cybersecurity: CyberGym 84.5%, top published result ahead of Mythos 5 (83.8%) and GPT-5.6 Sol (83.6%)
  • Long-horizon coding: DeepSWE v1.1 66.9% (from 46.2%), Terminal-Bench 3.0 28.3 (six-fold over 4.6)
  • Agentic automation: AutomationBench v1.0.6 jumps 26.2 to 48.2 (+84%); SAO reinforcement learning drives the long-horizon gains
  • Cost-sensitive frontier-adjacent coding: re-verified on Z.ai's official table at $1.40 input / $4.40 output per 1M, the same standing rate as GLM-5.2 and GLM-5.1 despite the post-training gains
  • Self-hosters: the open weights are now published on Hugging Face (zai-org/GLM-5.3, about 756 GB of safetensors) on the same GLM-5.2 base

Déconseillé pour

  • Deep offensive exploitation: ExploitBench 54.4% still trails Mythos 5 (78.0%) and GPT-5.6 Sol (76.5%)
  • Budgeting the Flash tier off the launch discount: the 50% promo is no longer on Z.ai's pricing table, which lists GLM-5.3-Flash at $0.15 input / $0.50 output per 1M
  • Modest self-hosting hardware: the published checkpoint is roughly 756 GB, a multi-GPU node deployment rather than a workstation one
  • Vision and multimodal workloads: no multimodal capability announced or benchmarked for GLM-5.3; Z.ai's vision line is a separate GLM-4.6V family
  • Buyers who need independently reproduced results: every headline benchmark remains vendor-run

Performances par tâche

Defensive security (CyberGym)

Excellent

84,5 % sur CyberGym — meilleur résultat publié, devant la frontière fermée ; rapporté par le fournisseur, non audité indépendamment

Long-horizon software engineering (DeepSWE v1.1)

Very Good

66,9 % sur DeepSWE v1.1, contre 46,2 % — un bond de 20 points, bien que le chiffre soit exécuté par Z.ai

Terminal/CLI coding (Terminal-Bench 3.0)

Very Good

28,3 sur Terminal-Bench 3.0, six fois plus que les 4,6 de GLM-5.2 ; au niveau de la frontière sur Terminal-Bench 2.1 (88,2 contre 88,8)

Deep exploitation (ExploitBench)

Fair

54,4 % sur ExploitBench, plus du double de GLM-5.2 (24,4 %), mais reste derrière la frontière fermée de plus de 20 points

Vision / multimodal

Poor

Aucune capacité de vision annoncée ni benchmarkée au lancement

Tarifs

Entrée

$1.40 / 1M tokens

Sortie

$4.40 / 1M tokens

Contexte

1M ctx; Flash $0.15/$0.50, launch promo ended

Voir tous les tarifs

Tests de performance

TestScoreSource
CyberGym84.5% Source
DeepSWE v1.166.9% Source
Terminal-Bench 3.028.3 Source
Terminal-Bench 2.188.2 Source
ExploitBench54.4% Source
ExploitGym (2h / 6h tasks)105 / 130 Source

Historique des verdicts

14 sept. 2026
conditional -> conditional (summary corrected): Rating unchanged: the headline coding and cyber benchmarks are still run by Z.ai's own team. The standing summary is corrected: it said per-token pricing was unpublished when Z.ai's table lists $1.40/$4.40, and the entity had stored the expired GLM-5.3-Flash launch promotion as if it were the list rate.

Sources

Journal de vérification

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

  • Tarifs· Aucun changement

    Agent automatisé

Comment nous évaluons