Grok-4 Heavy

Fiche modèle · xAI · grok-4-heavy

Grok 4 Heavy is the multi-agent version of Grok 4, released alongside the standard model in summer 2025. This system spawns multiple Grok 4 agents in parallel that work independently on problems and then collaborate by comparing their solutions, similar to a study group. The agents share insights and tricks they discover, with the system intelligently combining their work rather than simply using majority voting. Grok 4 Heavy uses approximately 10x more test-time compute than regular Grok 4, enabling it to solve significantly more complex problems. On the Humanities Last Exam, it achieves over 50% accuracy on text-only problems, and it scored a perfect result on the AIME 2025 mathematics competition. The system represents a major advancement in multi-agent AI collaboration and reasoning capabilities.

Identité

OrganisationxAI
Pays—
Famille—
Typellm
Date de sortie—
Licenceproprietary
Licence commercialenon
Licence open sourcenon
Poids ouvertsnon
Auto-hébergement (DGX Spark)non — poids fermés
Modalités—
Fenêtre de contexte—
KV cache / token (Q8)—
Paramètres totaux—
Paramètres actifs (session)—
Source des paramètres actifs—
Curation des paramètres actifs—
Prix par session—
Tokens / tâche (Data Players)1.2 M
Tokens / tâche (Terminal-Bench)83.3 M (est.)
Ratio Terminal-Bench / Data Players69.4×
Coût / tâche (Terminal-Bench)—
Tokens d'entraînement—
Coupure de connaissance2024-12-31
Créé le2026-05-07T16:53:39.919292Z
Mis à jour le2026-11-04T06:05:13.287124Z
SourceLLM Stats
URLhttps://llm-stats.com/models/grok-4-heavy

Prix par fournisseur

Toutes sources confondues : catalogue llm-stats, routeurs EUrouter/Cortecs et catalogues directs (OVHcloud, Scaleway, Infomaniak). Les prix sont convertis en USD pour la comparaison ; le prix publié et le taux BCE utilisé sont dans l'infobulle. Les prix inconnus restent indiqués « — ».

Aucune donnée.

Souveraineté qualifiée par offre (modèle × fournisseur) ; survolez le badge pour les critères.

Souveraineté

Scores par catégorie (top_scores)

CatégorieScore
math0.80
physics0.90
reasoning0.80
general0.80
biology0.90
chemistry0.90
code0.80
vision0.50

Scores par benchmark

BenchmarkCatégorieScoreMaxNormalisé (API)VérifiéAuto-déclaréDaté du
aime-2025math1.001.001.00nonoui2026-11-04T06:05:13.290816Z
gpqareasoning0.881.000.88nonoui2026-11-04T06:05:13.294218Z
hmmt25math0.971.000.97nonoui2026-11-04T06:05:13.297059Z
humanity's-last-examreasoning0.511.000.51nonoui2026-11-04T06:05:13.300943Z
livecodebenchreasoning0.791.000.79nonoui2026-11-04T06:05:13.304514Z
usamo25math0.621.000.62nonoui2026-11-04T06:05:13.307735Z

Classements

CatégorieRangScoreNote conservatriceBenchmarks évaluésMéthodeDaté duProvenance
math100.8041.694trueskill2026-11-04T08:08:42.442591Zrankings
code128—17.621trueskill2026-10-19model_history

Lecture : Note conservatrice = borne basse TrueSkill (niveau estimé − marge d’incertitude) ; Benchmarks évalués = nombre de benchmarks de la catégorie pris en compte. À performances égales, un modèle évalué sur moins de benchmarks obtient une note plus basse : la note ne se déduit pas d’un seul score. Provenance : rankings = top 50 de /rankings ; model_history = historique du modèle (/models/{id}/history), utilisé pour les rangs au-delà du top 50.

Données

Dates de curation (récupération depuis l'API) : models 2026-11-04 · benchmarks 2026-11-04 · scores 2026-11-04 · updates 2026-11-04 · account 2026-11-04 · model_history 2026-11-04 · rankings 2026-11-04