Capacité · 2026-11-12
Modèles d'IA à poids ouverts
Modèles publiant leurs poids entraînés sous licence — self-hosting, fine-tuning et environnements régulés.
Qu'est-ce que c'est ?
- Les modèles à poids ouverts publient leurs poids (souvent sous Apache 2.0, MIT ou une licence de recherche ouverte).
- Note : poids ouverts ≠ open source — les données d'entraînement et le code ne sont généralement pas publiés.
Pourquoi c'est important
- Self-hosting sur vos propres GPUs, fine-tuning sur données privées, fonctionnement hors ligne, déploiement en environnement régulé.
- Le tableau affiche l'hébergeur API le moins cher — avec votre propre matériel, le même modèle peut tourner sans coût d'API.
931 modèles avec cette capacité
Top 60 sur 931 affichés. Utilisez le répertoire complet pour filtrer davantage.
Frequently asked questions
How many AI models support poids ouverts?
931 canonical models in our database currently support poids ouverts. The list is regenerated on every data refresh, so it always reflects the latest releases tracked in our catalogue.
What is the cheapest model with poids ouverts?
Qwen3.8 27B from Alibaba (Qwen) is currently the lowest-priced option, at $0.001 per 1M input tokens and $0.001 per 1M output tokens. The full table above is sorted price-ascending.
Which model with poids ouverts has the largest context window?
Llama 4 Scout 17B Instruct (US) (Meta) leads on context at 10M tokens. This may matter if you also need long-document understanding alongside poids ouverts.
Which models are available on the most providers?
Production-readiness usually correlates with how many independent providers host the same weights. The top three by provider count are: GLM-5.3-Flash (92), GLM-5.3 (92), GLM-5.2 (90).
How is poids ouverts different from a regular LLM?
Open-weight models publish their trained weights publicly. You can self-host on your own GPUs, fine-tune on private data or run offline. Note: open weights ≠ open source — training data and code are usually not released.
How often is this list updated?
Daily. Our data pipeline syncs once a day, regenerates the canonical model list, and rebuilds these pages so newly released models appear within 24 hours.
Explore more
Top models with this capability
- Qwen3.8 27B$0.00 in / $0.00 out
- Ambient Small$0.00 in / $0.00 out
- Whisper Large v3$0.00 in / $0.00 out
- Voxtral Mini 3B$0.00 in / $0.00 out
- Multi-QA-mpnet-base-dot-v1$0.01 in / $0.00 out
Other capabilities
Best-of lists you might also want
Pricing comparisons
Vendors in this list
Dernière mise à jour :
Prices in USD per 1M tokens. Unknown means the provider does not publish per-token pricing.
Pricing and capabilities are refreshed daily and reconciled against each provider's official documentation. Always verify critical production decisions with the provider directly.