Qwen3-Next-80B-A3B-Thinking
Qwen3-Next-80B-A3B-Thinking is the thinking variant of the Qwen3-Next series, featuring the same groundbreaking architecture as the instruct model. Leveraging GSPO, it addresses stability and efficiency challenges of hybrid attention + high-sparsity MoE in RL training. It uses Hybrid Attention combining Gated DeltaNet and Gated Attention for efficient ultra-long context modeling, High-Sparsity MoE with 512 experts (10 activated + 1 shared), and Multi-Token Prediction. With 80B total parameters and only 3B activated, it demonstrates outstanding performance on complex reasoning tasks — outperforming Qwen3-30B-A3B-Thinking-2507, Qwen3-32B-Thinking, and even the proprietary Gemini-2.5-Flash-Thinking across multiple benchmarks. Architecture: 48 layers, 15T training tokens, hybrid layout of 12*(3*(Gated DeltaNet->MoE)->(Gated Attention->MoE)). Supports only thinking mode with automatic <think> tag inclusion, may generate longer thinking content.
Identité
| Organisation | Alibaba Cloud / Qwen Team |
|---|---|
| Pays | — |
| Famille | — |
| Type | llm |
| Date de sortie | 2025-09-10 |
| Licence | apache 2.0 |
| Licence commerciale | oui |
| Licence open source | oui |
| Poids ouverts | oui |
| Auto-hébergement (DGX Spark) | oui — ≤ 107 Md (Q8/FP8) |
| Modalités | text |
| Fenêtre de contexte | 262 144 |
| KV cache / token (Q8) | 49.2 ko/token |
| Paramètres totaux | 80 B |
| Paramètres actifs (session) | 3 B |
| Source des paramètres actifs | actifs : nom (heuristique) ; total : config Hugging Face (Qwen/Qwen3-Next-80B-A3B-Thinking) |
| Curation des paramètres actifs | 2026-11-04 |
| Prix par session | — |
| 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) | — |
| Tokens d'entraînement | 15 000 000 000 000 |
| Coupure de connaissance | — |
| Créé le | 2026-05-07T16:53:36.358328Z |
| Mis à jour le | 2026-11-04T06:05:06.729453Z |
| Source | LLM Stats |
| URL | https://llm-stats.com/models/qwen3-next-80b-a3b-thinking |
IA Act — preuves fournisseur
| Critère | Note / 5 | Poids |
|---|---|---|
| Documentation technique | 2,5 | 5 % |
| Résumé des données d'entraînement | 2,5 | 35 % |
| Conformité droit d'auteur | 2,5 | 35 % |
| Code of Practice GPAI | 1,2 | 20 % |
| Risque systémique | 0,0 | 5 % |
Curation : 2026-10-23.
IA ActPrix par fournisseur
Statut du modèle (offres + auto-hébergement) : partiel — auto-hébergement 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.
SouverainetéScores par catégorie (top_scores)
| Catégorie | Score |
|---|---|
| chat | 0.70 |
| instruction_following | 0.80 |
| language | 0.80 |
| legal | 0.80 |
| math | 0.80 |
| multimodal | 0.60 |
| physics | 0.70 |
| reasoning | 0.70 |
| spatial_reasoning | 0.60 |
| structured_output | 0.80 |
| finance | 0.80 |
| general | 0.70 |
| healthcare | 0.70 |
| agents | 0.70 |
| biology | 0.80 |
| chemistry | 0.70 |
| communication | 0.60 |
| creativity | 0.70 |
| economics | 0.60 |
| tool_calling | 0.60 |
| vision | 0.60 |
| writing | 0.70 |
Scores par benchmark
| Benchmark | Catégorie | Score | Max | Normalisé (API) | Vérifié | Auto-déclaré | Daté du |
|---|---|---|---|---|---|---|---|
| aime-2025 | math | 0.88 | 1.00 | 0.88 | non | oui | 2026-11-04T06:05:06.745540Z |
| arena-hard-v2 | general | 0.62 | 1.00 | 0.62 | non | oui | 2026-11-04T06:05:06.767026Z |
| bfcl-v3 | general | 0.72 | 1.00 | 0.72 | non | oui | 2026-11-04T06:05:06.774358Z |
| cfeval | code | 2 071.00 | 10 000.00 | 0.21 | non | oui | 2026-11-04T06:05:06.757556Z |
| gpqa | reasoning | 0.77 | 1.00 | 0.77 | non | oui | 2026-11-04T06:05:06.738735Z |
| hmmt25 | math | 0.74 | 1.00 | 0.74 | non | oui | 2026-11-04T06:05:06.748681Z |
| ifeval | chat | 0.89 | 1.00 | 0.89 | non | oui | 2026-11-04T06:05:06.763564Z |
| include | general | 0.79 | 1.00 | 0.79 | non | oui | 2026-11-04T06:05:06.801340Z |
| livebench-20241125 | math | 0.77 | 1.00 | 0.77 | non | oui | 2026-11-04T06:05:06.751589Z |
| livecodebench-v6 | reasoning | 0.69 | 1.00 | 0.69 | non | oui | 2026-11-04T06:05:06.754503Z |
| mmlu-pro | language | 0.83 | 1.00 | 0.83 | non | oui | 2026-11-04T06:05:06.732823Z |
| mmlu-prox | language | 0.79 | 1.00 | 0.79 | non | oui | 2026-11-04T06:05:06.797861Z |
| mmlu-redux | language | 0.93 | 1.00 | 0.93 | non | oui | 2026-11-04T06:05:06.735802Z |
| multi-if | chat | 0.78 | 1.00 | 0.78 | non | oui | 2026-11-04T06:05:06.794907Z |
| ojbench | reasoning | 0.30 | 1.00 | 0.30 | non | oui | 2026-11-04T06:05:06.760493Z |
| polymath | math | 0.56 | 1.00 | 0.56 | non | oui | 2026-11-04T06:05:06.804625Z |
| supergpqa | reasoning | 0.61 | 1.00 | 0.61 | non | oui | 2026-11-04T06:05:06.741898Z |
| tau-bench-airline | reasoning | 0.49 | 1.00 | 0.49 | non | oui | 2026-11-04T06:05:06.780556Z |
| tau-bench-retail | chat | 0.70 | 1.00 | 0.70 | non | oui | 2026-11-04T06:05:06.777431Z |
| tau2-airline | chat | 0.60 | 1.00 | 0.60 | non | oui | 2026-11-04T06:05:06.787600Z |
| tau2-retail | communication | 0.68 | 1.00 | 0.68 | non | oui | 2026-11-04T06:05:06.783965Z |
| tau2-telecom | communication | 0.44 | 1.00 | 0.44 | non | oui | 2026-11-04T06:05:06.791136Z |
| writingbench | writing | 0.85 | 1.00 | 0.85 | non | oui | 2026-11-04T06:05:06.770534Z |
Classements
| Catégorie | Rang | Score | Note conservatrice | Benchmarks évalués | Méthode | Daté du | Provenance |
|---|---|---|---|---|---|---|---|
| chat | 31 | 0.70 | 22.86 | 4 | trueskill | 2026-11-04T08:08:43.789169Z | rankings |
| agents | 117 | — | 12.18 | 1 | 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