Hy4 preview

Fiche modèle · Tencent · hy4-preview

Hy4 preview is a new-generation Mixture-of-Experts (MoE) flagship model from the Tencent Hy Team (released 2026-08-28). It has 770B total parameters with 49B activated per token, plus a native MTP layer (10B total / 0.7B activated) for speculative decoding. The backbone uses 78 layers (first dense FFN, then 77 MoE layers with 256 routed experts + 1 shared expert, top-8 routing), Gated DeepSeek Sparse Attention with IndexCache, and identity Hyper-Connections, with a 1,048,576-token context window. It is text-in / text-out, Apache 2.0 licensed, and a hybrid-thinking model defaulting to high reasoning effort (also supports no_think). Training targets productivity workloads — software engineering, office/analysis, game development, and scientific research — and is co-designed with Tencent products such as CodeBuddy and WorkBuddy. As an early preview, known limitations include longer-than-necessary reasoning on complex tasks and a tendency to over-verify its own work.

Identité

OrganisationTencent
Pays—
Famille—
Typellm
Date de sortie2026-08-28
Licenceapache 2.0
Licence commercialeoui
Licence open sourceoui
Poids ouvertsoui
Auto-hébergement (DGX Spark)non — 770 Md > 107 Md
Modalités—
Fenêtre de contexte1 048 576
KV cache / token (Q8)79.9 ko/token
Paramètres totaux770 B
Paramètres actifs (session)49 B
Source des paramètres actifsweb : https://huggingface.co/tencent/Hy4-preview
Curation des paramètres actifs2026-11-04
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 connaissance—
Créé le2026-08-31T00:25:44.033159Z
Mis à jour le2026-11-04T06:05:12.691070Z
SourceLLM Stats
URLhttps://llm-stats.com/models/hy4-preview

Prix par fournisseur

Statut du modèle (offres + auto-hébergement) : non — hors DGX Spark (> 107 Md).

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
knowledge0.70
legal1 678.00
long_context0.70
math0.40
physics0.90
productivity0.60
reasoning67.70
safety0.80
science0.70
search0.80
finance839.40
general53.00
agents76.80
biology0.80
chemistry0.80
code0.60
systems0.40
tool_calling0.60

Scores par benchmark

BenchmarkCatégorieScoreMaxNormalisé (API)VérifiéAuto-déclaréDaté du
agents-last-examagents0.231.000.23nonoui2026-11-04T06:05:12.759178Z
apex-agentsreasoning0.371.000.37nonoui2026-11-04T06:05:12.746762Z
arxivmathmath0.671.000.67nonoui2026-11-04T06:05:12.790301Z
automationbenchagents0.321.000.32nonoui2026-11-04T06:05:12.766998Z
bankertoolbenchagents0.791.000.79nonoui2026-11-04T06:05:12.769647Z
biomysterybenchbiology0.711.000.71nonoui2026-11-04T06:05:12.772719Z
critptreasoning0.171.000.17nonoui2026-11-04T06:05:12.778314Z
cybergymcode0.781.000.78nonoui2026-11-04T06:05:12.725547Z
deepswecode0.641.000.64nonoui2026-11-04T06:05:12.703446Z
dracoagents0.771.000.77nonoui2026-11-04T06:05:12.733648Z
gdpval-aageneral1 678.003 000.000.56nonoui2026-11-04T06:05:12.764018Z
gpqareasoning0.921.000.92nonoui2026-11-04T06:05:12.780999Z
horizonmathmath0.091.000.09nonoui2026-11-04T06:05:12.793585Z
humanity's-last-exam-(no-tools,-text-only)reasoning0.431.000.43nonoui2026-11-04T06:05:12.783859Z
humanity's-last-exam-(with-tools,-text-only)reasoning0.551.000.55nonoui2026-11-04T06:05:12.775609Z
job-benchagents0.621.000.62nonoui2026-11-04T06:05:12.753737Z
matharena-apexmath0.741.000.74nonoui2026-11-04T06:05:12.796966Z
mcp-atlasreasoning0.841.000.84nonoui2026-11-04T06:05:12.736225Z
nl2repocode0.591.000.59nonoui2026-11-04T06:05:12.707543Z
officeqa-proagents0.661.000.66nonoui2026-11-04T06:05:12.744055Z
onemillion-benchagents0.651.000.65nonoui2026-11-04T06:05:12.731053Z
posttrainbenchagents0.361.000.36nonoui2026-11-04T06:05:12.712504Z
program-benchagents0.171.000.17nonoui2026-11-04T06:05:12.719495Z
skillsbenchcode0.631.000.63nonoui2026-11-04T06:05:12.750377Z
superchemchemistry0.661.000.66nonoui2026-11-04T06:05:12.787291Z
swe-bench-multilingualreasoning0.831.000.83nonoui2026-11-04T06:05:12.694932Z
swe-bench-proreasoning0.661.000.66nonoui2026-11-04T06:05:12.699226Z
swe-marathoncode0.321.000.32nonoui2026-11-04T06:05:12.716799Z
terminal-bench-2.1reasoning0.851.000.85nonoui2026-11-04T06:05:12.722652Z
toolathlonagents0.741.000.74nonoui2026-11-04T06:05:12.739943Z
widesearchsearch0.841.000.84nonoui2026-11-04T06:05:12.728242Z
workspace-benchagents0.601.000.60nonoui2026-11-04T06:05:12.756414Z

Classements

CatégorieRangScoreNote conservatriceBenchmarks évaluésMéthodeDaté duProvenance
agents3276.8034.6222trueskill2026-11-04T08:08:47.839150Zrankings
code310.6037.2211trueskill2026-11-04T08:08:39.742311Zrankings
knowledge150.7015.491trueskill2026-11-04T08:08:45.139626Zrankings
math270.4038.396trueskill2026-11-04T08:08:42.442591Zrankings
reasoning1967.7049.0525trueskill2026-11-04T08:08:41.093494Zrankings

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