Model Index
The leading open-weight models: origin, licence and hardware. Updated monthly.
Europe
Mistral leads. Most licences are permissive; Mistral Medium 3.5 is the exception for large companies.
Model | Released | Size | Licence | What it means for you | Minimum hardware |
|---|---|---|---|---|---|
| Mistral Small 4 Mistral AI (FR) | Mar 2026 | 119B MoE, 6B active | Apache 2.0 | No commercial restrictions | 4× H100 or 2× H200 |
| Mistral Large 3 Mistral AI (FR) | Dec 2025 | 675B MoE, 41B active | Apache 2.0 | No commercial restrictions | 8× H200 (FP8) |
| Mistral Medium 3.5 Mistral AI (FR) | Apr 2026 | 128B dense | Modified MIT | No rights above $20M monthly revenue without a commercial licence | 2× H200 (FP8) |
| EuroLLM-22B EU research consortium | Dec 2025 | 22B dense | Apache 2.0 | No commercial restrictions | 1 GPU |
| TildeOpen-30B Tilde (LV) | 2025 | 30B dense | CC BY 4.0 | Attribution required | 1 GPU |
United States
Strong permissive options from OpenAI, Google and IBM. Meta's Llama licence limits EU companies.
Model | Released | Size | Licence | What it means for you | Minimum hardware |
|---|---|---|---|---|---|
| gpt-oss-120b OpenAI | Aug 2025 | 117B MoE | Apache 2.0 | No commercial restrictions | 1× 80 GB GPU |
| Gemma 4 Google | Apr 2026 | Up to 31B | Apache 2.0 | No commercial restrictions | 1 GPU |
| Nemotron 3 Ultra NVIDIA | Jun 2026 | 550B MoE, 55B active | NVIDIA open model licence | Custom terms; review before use | 2 to 4× B200 |
| Llama 4 Maverick Meta | Apr 2025 | 400B MoE, 17B active | Llama 4 Community | Multimodal rights withheld from EU-based companies | 8× H100 |
| Granite 4.2 IBM | Aug 2026 | 3B to 30B | Apache 2.0 | ISO 42001-certified model family | 1 GPU |
China
The strongest open models today. Custom licences at the top end; origin is a policy question.
Model | Released | Size | Licence | What it means for you | Minimum hardware |
|---|---|---|---|---|---|
| MiMo-V2.6-Pro Xiaomi | Sep 2026 | 1.02T MoE, 42B active | MIT | No commercial restrictions | 8× B200 |
| DeepSeek V4 Pro DeepSeek | Apr 2026 | 1.6T MoE, 49B active | MIT | No commercial restrictions | 8× H200 or 8× B300 |
| Kimi K3 Moonshot AI | Jul 2026 | 2.8T MoE, 104B active | Kimi K3 License | Internal use allowed; resale as a service limited above $20M revenue | 8× B200 minimum |
| Qwen3.8 (flagship) Alibaba | Aug 2026 | 2.4T MoE, 95B active | Qwen3.8-Max License | Internal use allowed; smaller Qwen models are Apache 2.0 | Multi-node |
| GLM-5.3 Z.ai | Aug 2026 | About 750B | GLM-5.3 License | Custom terms; Flash variant is MIT | 8× H200 |
Licence legend
- Permissive No material restrictions for enterprise use.
- Conditional Custom licence. Review the terms before use.
- Restricted Excludes large or EU-based companies from some rights.
Sources: model cards and licence texts on Hugging Face and vendor sites, accessed 27 September 2026. Hardware is indicative for production inference at 8-bit or 4-bit precision. Licence summaries are not legal advice.