DeepSeek-V3.2
DeepSeek-V3.2 is a 685B-parameter MoE model that harmonizes high computational efficiency with superior reasoning and agent performance. It introduces DeepSeek Sparse Attention (DSA) for efficient long-context processing, a scalable reinforcement learning post-training framework, and large-scale agentic task synthesis covering 1,800+ environments. V3.2 achieves GPT-5-level performance across reasoning, coding, and agentic benchmarks, with gold-medal results from its Speciale variant on IMO, IOI, ICPC World Finals, and CMO 2025.
Identité
| Organisation | DeepSeek |
|---|---|
| Pays | — |
| Famille | — |
| Type | llm |
| Date de sortie | 2025-12-01 |
| Licence | mit |
| Licence commerciale | oui |
| Licence open source | oui |
| Poids ouverts | oui |
| Auto-hébergement (DGX Spark) | non — 685 Md > 107 Md |
| Modalités | text |
| Fenêtre de contexte | 163 840 |
| KV cache / token (Q8) | 874.5 ko/token |
| Paramètres totaux | 685 B |
| Paramètres actifs (session) | 37 B |
| Source des paramètres actifs | web : https://huggingface.co/deepseek-ai/DeepSeek-V3.2 |
| Curation des paramètres actifs | 2026-11-04 |
| Prix par session | 0.34 $ (1 000 000 tokens entrée + 200 000 sortie, prix le moins cher par fournisseur) |
| Tokens / tâche (Data Players) | 1.2 M |
| Tokens / tâche (Terminal-Bench) | 83.3 M (est.) |
| Ratio Terminal-Bench / Data Players | 69.4× |
| Coût / tâche (Terminal-Bench) | 2.82 $ (est.) |
| Tokens d'entraînement | — |
| Coupure de connaissance | — |
| Créé le | 2026-05-07T16:53:28.330796Z |
| Mis à jour le | 2026-11-04T06:04:39.969311Z |
| Source | LLM Stats |
| URL | https://llm-stats.com/models/deepseek-v3.2 |
IA Act — preuves fournisseur
| Critère | Note / 5 | Poids |
|---|---|---|
| Documentation technique | 2,5 | 5 % |
| Résumé des données d'entraînement | 5,0 | 35 % |
| Conformité droit d'auteur | 5,0 | 35 % |
| Code of Practice GPAI | 3,8 | 20 % |
| Risque systémique | 0,0 | 5 % |
Curation : 2026-10-23.
IA ActPrix par fournisseur
Statut du modèle (offres + auto-hébergement) : souverain — via Arkane (cortecs), Tensorix (eurouter).
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 « — ».
| Fournisseur | Source | Souveraineté | Entrée (/M tok) | Sortie (/M tok) | Quantification | Contexte | Latence (s) | Débit (tok/s) | Statut |
|---|---|---|---|---|---|---|---|---|---|
| Arkane | cortecs | souverain | 0.50 $ | 1.51 $ | — | 163 840 | — | — | — |
| Tensorix | eurouter | souverain | — | — | — | 163 840 | — | — | — |
| AWS Bedrock | eurouter | partiel | — | — | — | 163 840 | — | — | — |
| DeepInfra | llm-stats | non | 0.26 $ | 0.38 $ | — | — | — | — | active |
Souveraineté qualifiée par offre (modèle × fournisseur) ; survolez le badge pour les critères.
SouverainetéScores par catégorie (top_scores)
| Catégorie | Score |
|---|---|
| language | 0.80 |
| legal | 0.80 |
| math | 0.80 |
| physics | 0.80 |
| reasoning | 0.70 |
| search | 0.60 |
| finance | 0.80 |
| frontend_development | 0.70 |
| general | 0.70 |
| healthcare | 0.80 |
| agents | 0.50 |
| biology | 0.80 |
| chemistry | 0.80 |
| code | 0.70 |
| tool_calling | 0.50 |
| vision | 0.40 |
Scores par benchmark
| Benchmark | Catégorie | Score | Max | Normalisé (API) | Vérifié | Auto-déclaré | Daté du |
|---|---|---|---|---|---|---|---|
| aime-2025 | math | 0.93 | 1.00 | 0.93 | non | oui | 2026-11-04T06:04:39.991622Z |
| browsecomp | reasoning | 0.51 | 1.00 | 0.51 | non | oui | 2026-11-04T06:04:40.116724Z |
| browsecomp-zh | reasoning | 0.65 | 1.00 | 0.65 | non | oui | 2026-11-04T06:04:40.129425Z |
| codeforces | math | 0.80 | 3 000.00 | 0.80 | non | oui | 2026-11-04T06:04:40.053128Z |
| gpqa | reasoning | 0.82 | 1.00 | 0.82 | non | oui | 2026-11-04T06:04:39.987946Z |
| hmmt-2025 | math | 0.90 | 1.00 | 0.90 | non | oui | 2026-11-04T06:04:40.019368Z |
| humanity's-last-exam | reasoning | 0.41 | 1.00 | 0.41 | non | oui | 2026-11-04T06:04:40.133093Z |
| imo-answerbench | math | 0.78 | 1.00 | 0.78 | non | oui | 2026-11-04T06:04:40.034671Z |
| livecodebench | reasoning | 0.83 | 1.00 | 0.83 | non | oui | 2026-11-04T06:04:40.063204Z |
| mcp-mark | agents | 0.38 | 1.00 | 0.38 | non | oui | 2026-11-04T06:04:40.101904Z |
| mcp-universe | agents | 0.46 | 1.00 | 0.46 | non | oui | 2026-11-04T06:04:40.096838Z |
| mmlu-pro | language | 0.85 | 1.00 | 0.85 | non | oui | 2026-11-04T06:04:39.983722Z |
| swe-bench-multilingual | reasoning | 0.70 | 1.00 | 0.70 | non | oui | 2026-11-04T06:04:40.078717Z |
| swe-bench-verified | reasoning | 0.73 | 1.00 | 0.73 | non | oui | 2026-11-04T06:04:40.073130Z |
| t2-bench | agents | 0.80 | 1.00 | 0.80 | non | oui | 2026-11-04T06:04:40.091784Z |
| terminal-bench-2 | reasoning | 0.46 | 1.00 | 0.46 | non | oui | 2026-11-04T06:04:40.083759Z |
| toolathlon | agents | 0.35 | 1.00 | 0.35 | non | oui | 2026-11-04T06:04:40.106704Z |
Classements
| Catégorie | Rang | Score | Note conservatrice | Benchmarks évalués | Méthode | Daté du | Provenance |
|---|---|---|---|---|---|---|---|
| agents | 119 | — | 11.99 | 5 | trueskill | 2026-10-19 | model_history |
| code | 89 | — | 22.76 | 4 | trueskill | 2026-10-19 | model_history |
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