License-first guide

Open-Weight Models for Commercial Use

Weight access is not a commercial-use license.

This guide separates standard permissive licenses from special and conditional model terms across the 32 verified Model Passports. It is designed for technical screening before a formal legal review.

Apache 2.0Commercial use broadly permitted with license/notice obligations.
MITShort permissive license with notice preservation.
Modified / customRead thresholds, attribution, redistribution and geography.
Provider termsManaged APIs can add a separate contract layer.
The key distinction

Commercial use is a license field, not an “open” synonym.

Two models can both publish downloadable weights and still have materially different commercial conditions.

Apache 2.0 and MIT are standard permissive licenses represented heavily in this registry. Other releases use custom terms. Llama Community licenses include model-specific redistribution and scale clauses; Gemma uses its own Terms of Use; Qwen2.5-VL uses a Qwen license; Kimi K2 uses a modified MIT license with an additional scale-triggered branding provision.

OpenWeightModels therefore stores commercial use separately from weight availability, modification rights, redistribution, incorporated use policies and deployment-provider terms. A green “commercial use” field should never be treated as the complete legal analysis.

Standard permissive

23 models recorded as commercially allowed.

This group is made up of exact Apache 2.0 and MIT model records in the registry. Separate policies and application law can still matter.

ModelLicenseCommercialQualificationPassport
DeepSeek-R1DeepSeekMITAllowedShort permissive license. It does not itself impose a use-field restriction; application law, provider terms and separate policies still apply.Review →
DeepSeek-V3.1DeepSeekMITAllowedShort permissive license. It does not itself impose a use-field restriction; application law, provider terms and separate policies still apply.Review →
Devstral Small 2505Mistral AIApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
GLM-4.5Z.aiMITAllowedShort permissive license. It does not itself impose a use-field restriction; application law, provider terms and separate policies still apply.Review →
GLM-4.5 AirZ.aiMITAllowedShort permissive license. It does not itself impose a use-field restriction; application law, provider terms and separate policies still apply.Review →
gpt-oss-120bOpenAIApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
gpt-oss-20bOpenAIApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Granite 3.3 8B InstructIBMApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Magistral Small 2506Mistral AIApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Mathstral 7B v0.1Mistral AIApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Mistral Nemo Instruct 2407Mistral AI / NVIDIAApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Mistral Small 3.1 24B InstructMistral AIApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
OLMo 2 13B InstructAi2Apache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
OLMo 2 32B InstructAi2Apache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Phi-4MicrosoftMITAllowedShort permissive license. It does not itself impose a use-field restriction; application law, provider terms and separate policies still apply.Review →
Phi-4 Mini InstructMicrosoftMITAllowedShort permissive license. It does not itself impose a use-field restriction; application law, provider terms and separate policies still apply.Review →
Phi-4 Multimodal InstructMicrosoftMITAllowedShort permissive license. It does not itself impose a use-field restriction; application law, provider terms and separate policies still apply.Review →
Qwen2.5-Coder-32B-InstructQwen / AlibabaApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Qwen3-235B-A22BQwen / AlibabaApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Qwen3-30B-A3BQwen / AlibabaApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Qwen3-32BQwen / AlibabaApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Qwen3-Coder-480B-A35B-InstructQwen / AlibabaApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
SmolLM3 3BHugging FaceApache 2.0AllowedIncludes an express patent grant. Model-specific acceptable-use policies, if separately published, are not automatically part of Apache 2.0.Review →
Special condition

Commercial use with an additional trigger.

Modified permissive licenses can preserve broad rights while adding model-specific conditions.

ModelLicenseStatusSpecial conditionPassport
Kimi K2 InstructMoonshot AIModified MITAllowed with special conditionIf a commercial product/service exceeds either 100M monthly active users or $20M monthly revenue, the published modified license requires prominent “Kimi K2” display in the product/service user interface.Review →
Conditional custom terms

8 models require a term-by-term review.

Conditional does not mean “non-commercial.” It means a binary yes/no would hide material requirements such as distribution obligations, scale conditions, region restrictions or custom model-use terms.

ModelLicenseStatusCondition summaryPassport
Gemma 3 12B ITGoogle DeepMindGemma TermsConditionalThe Gemma Terms define distribution broadly enough to include making Gemma or derivatives available through hosted services. Recipients, notices, modified-file marking and prohibited-use obligations should be reviewed before redistribution.Review →
Gemma 3 27B ITGoogle DeepMindGemma TermsConditionalThe Gemma Terms define distribution broadly enough to include making Gemma or derivatives available through hosted services. Recipients, notices, modified-file marking and prohibited-use obligations should be reviewed before redistribution.Review →
Gemma 3 4B ITGoogle DeepMindGemma TermsConditionalThe Gemma Terms define distribution broadly enough to include making Gemma or derivatives available through hosted services. Recipients, notices, modified-file marking and prohibited-use obligations should be reviewed before redistribution.Review →
Gemma 3n E4B ITGoogle DeepMindGemma TermsConditionalThe Gemma Terms define distribution broadly enough to include making Gemma or derivatives available through hosted services. Recipients, notices, modified-file marking and prohibited-use obligations should be reviewed before redistribution.Review →
Llama 3.3 70B InstructMetaLlama 3.3 CommunityConditionalCommercial and research use is intended, but the Community License and Acceptable Use Policy remain part of the deployment decision. Review the current official agreement for scale, attribution and redistribution conditions.Review →
Llama 4 MaverickMetaLlama 4 CommunityconditionalIf licensee/affiliate products or services exceeded 700M MAU in the calendar month preceding the Llama 4 release date, a separate license from Meta is required before exercising rights.Review →
Llama 4 ScoutMetaLlama 4 CommunityConditionalIncludes the 700M-MAU special licensing condition; the incorporated Llama 4 AUP contains a specific EU developer restriction for multimodal Llama 4 models while excluding end users of products/services from that particular restriction.Review →
Qwen2.5-VL-72B-InstructQwen / AlibabaQwen LicenseConditionalThe official Qwen license requires a separate license for specified commercial products/services above 100M monthly active users and requires “Built with Qwen” or “Improved using Qwen” attribution when Qwen materials/outputs are used to create or improve a distributed AI model.Review →
FAQ

Commercial-use questions.

Which open-weight models allow commercial use?

In the current 32-model registry, 23 models are recorded as commercially allowed under Apache 2.0 or MIT, one is allowed with a special condition, and eight use conditional custom terms. The exact license record should be checked before deployment.

Does open weight mean commercial use is allowed?

No. Weight availability and commercial permission are separate attributes. Custom licenses can add attribution, scale, redistribution, geography or acceptable-use conditions.

Are Apache 2.0 and MIT the simplest commercial licenses in the registry?

They are the main standard permissive licenses represented here. They generally permit commercial use, modification and redistribution subject to their license conditions.

Is this commercial-use page legal advice?

No. It is a technical comparison of recorded license terms. The official license, incorporated policies, provider terms and applicable law remain authoritative.