Open Systems, Blockchain & Business Daily

RSS ↗
OSBBD

AI, money and the software behind them.

Menu
OSBBDSearch

Open-weight AI models: licenses, sizes and hardware

A register of open-weight language models: models whose trained weights anyone can download from the developer’s official repository and run on their own hardware. For each one it lists the size, the active parameters for mixture-of-experts models, the context window, the license and whether commercial use is allowed, and how much space the official weights take.

45 listings

NameDeveloperReleasedTotal parametersActive parameters per tokenContext windowLicenseCommercial use
DeepSeek-V4.1-FlashDeepSeekSeptember 10, 2026552B + 196B memory8B–16B1MMITYes
GLM-5.3-FlashZ.ai (Zhipu AI)August 26, 2026320B18BMITYes
Qwen3.8-Flash-NextAlibaba QwenAugust 26, 2026125B6B256KQwen Community License 1.0Yes, with conditions
Granite 4.2 8BIBMAugust 25, 20268B128KApache 2.0Yes
Granite 4.2 30BIBMAugust 25, 202630B128KApache 2.0Yes
GLM-5.3Z.ai (Zhipu AI)August 14, 2026744B40BGLM-5.3 LicenseYes, with conditions
Qwen3.8-27BAlibaba QwenAugust 14, 202627B256KApache 2.0Yes
DeepSeek-V4-Pro-0813DeepSeekAugust 13, 20261.6T49B1MMITYes
Qwen3.8-2.4T-A95BAlibaba QwenAugust 12, 20262.4T95B256KQwen3.8-Max LicenseYes, with conditions
Nemotron 3.5 LightningNVIDIAAugust 11, 202630B3B1MOpenMDW License Agreement v1.1Yes
Kimi K3Moonshot AIJuly 20262.8T104B1MKimi K3 LicenseYes, with conditions
Nemotron 3 UltraNVIDIAJune 4, 2026550B55B1MOpenMDW License Agreement v1.1Yes
Gemma 4 12BGoogle DeepMindJune 3, 202611.95B256KApache 2.0Yes
MiniMax-M3MiniMaxMay 31, 2026~428B~23B1MMiniMax Community LicenseYes, with conditions
Mistral Medium 3.5Mistral AIMay 22, 2026128B256KModified MIT LicenseYes, with conditions
Command A+CohereMay 20, 2026218B25B128KApache 2.0Yes
Kimi K2.6Moonshot AIApril 21, 20261T32B256KModified MIT LicenseYes, with conditions
Qwen3.6-35B-A3BAlibaba QwenApril 16, 202635B3B256KApache 2.0Yes
Gemma 4 26B A4BGoogle DeepMindApril 2, 202625.2B3.8B256KApache 2.0Yes
Gemma 4 31BGoogle DeepMindApril 2, 202630.7B256KApache 2.0Yes
Gemma 4 E2BGoogle DeepMindApril 2, 20265.1B128KApache 2.0Yes
Gemma 4 E4BGoogle DeepMindApril 2, 20268B128KApache 2.0Yes
MiniMax-M2.7MiniMaxMarch 18, 2026229BMiniMax Non-Commercial LicenseNo
Mistral Small 4Mistral AIMarch 16, 2026119B6.5B256KApache 2.0Yes
Nemotron 3 SuperNVIDIAMarch 11, 2026120B12B1MNVIDIA Nemotron Open Model LicenseYes
Phi-4-reasoning-vision-15BMicrosoftMarch 4, 202615B16KMITYes
Qwen3.5-397B-A17BAlibaba QwenFebruary 16, 2026397B17B256KApache 2.0Yes
Qwen3-Coder-NextAlibaba QwenFebruary 3, 202680B3B256KApache 2.0Yes
Olmo 3.1 32BAllen Institute for AI (Ai2)December 12, 202532B65KApache 2.0Yes
Devstral 2Mistral AIDecember 9, 2025123B256KModified MIT LicenseYes, with conditions
Devstral Small 2Mistral AIDecember 9, 202524B256KApache 2.0Yes
Ministral 3 14BMistral AIDecember 2, 202514B256KApache 2.0Yes
Mistral Large 3Mistral AIDecember 2, 2025675B41B256KApache 2.0Yes
DeepSeek-V3.2DeepSeekDecember 1, 2025671B37BMITYes
Magistral Small 1.2Mistral AISeptember 202524B128KApache 2.0Yes
gpt-oss-20bOpenAIAugust 5, 202521B3.6B128KApache 2.0Yes
gpt-oss-120bOpenAIAugust 5, 2025117B5.1B128KApache 2.0Yes
DeepSeek-R1-0528DeepSeekMay 28, 2025671B37B128KMITYes
Llama 4 MaverickMetaApril 5, 2025400B17B1MLlama 4 Community LicenseYes, with conditions
Llama 4 ScoutMetaApril 5, 2025109B17B10MLlama 4 Community LicenseYes, with conditions
Command ACohereMarch 2025111B256KCC BY-NC 4.0No
Phi-4MicrosoftDecember 12, 202414B16KMITYes
Llama 3.3 70BMetaDecember 6, 202470B128KLlama 3.3 Community LicenseYes, with conditions
Llama 3.1 8BMetaJuly 23, 20248B128KLlama 3.1 Community LicenseYes, with conditions
Llama 3.1 405BMetaJuly 23, 2024405B128KLlama 3.1 Community LicenseYes, with conditions

“Open-weight” is not the same as “open-source”. Many of these licenses add conditions, such as a separate agreement above a revenue or user threshold, a required attribution line, or a ban on commercial use. The license columns summarize those conditions; the full license text is linked on each model’s page. For the background, read Open-weight vs. open-source AI.

How to read the sizes. Total parameters decide how much memory a model needs, because all of them have to be loaded, even in a mixture-of-experts model that uses only a fraction for each token. Active parameters decide speed. As a rough guide, a model takes about 2 GB per billion parameters in BF16, 1 GB in 8-bit and a little over 0.5 GB in 4-bit, before the context cache. The memory calculator checks which models fit your GPU and RAM, and Running AI locally explains the trade-offs.

The table covers major releases from Meta, Google DeepMind, OpenAI, Alibaba Qwen, DeepSeek, Mistral AI, Microsoft, NVIDIA, IBM, Moonshot AI, Z.ai, MiniMax, Ai2 and Cohere, and is checked against official sources every week.

OSBBD

What are you looking for?

Search by headline, topic or keyword.