induwara.lk
induwara.lkDevelopers · AI

AI LLM License & Commercial-Use Checker

Find out in seconds whether you can legally use Llama, Mistral, Gemma, Qwen, DeepSeek, Command R, GPT, Claude or Gemini in a commercial product — and the exact strings attached. Each verdict links the official license. No signup, no ads, informational only.

By Induwara AshinsanaUpdated Jun 19, 2026
Check an LLM's commercial license

Informational summaries of public license terms — not legal advice. Always read the linked official license before you ship. Terms change; check the "verified" date below.

Matches model name, family, vendor, and license. Leave empty to see everything.

Commercial use
Weights
Family
Allowed
10
Conditional
3
Not allowed
2
Via API
3

Showing 18 of 18 models

Mistral 7B / Mixtral 8x7B / 8x22B

Mistral AI · 7B · 8x7B · 8x22B

Allowed

Fully free for commercial use under Apache 2.0 — no MAU cap.

Apache 2.0Open weightsShip the license copy
Official license

Mistral Small 3

Mistral AI · 24B

Allowed

Apache 2.0 — commercial use allowed, no MAU cap, no attribution beyond the notice file.

Apache 2.0Open weightsShip the license copy
Official license

Qwen3 (0.6B–32B)

Alibaba Cloud · 0.6B–32B

Allowed

Apache 2.0 for these sizes — commercial use allowed, no MAU cap.

Apache 2.0Open weightsShip the license copy
Official license

DeepSeek-R1

DeepSeek · 671B (MoE)

Allowed

MIT — commercial use allowed with no MAU cap; just keep the license notice.

MIT LicenseOpen weights
Official license

DeepSeek-V3

DeepSeek · 671B (MoE)

Allowed

MIT-licensed weights — commercial use allowed, no MAU cap.

MIT LicenseOpen weights
Official license

Microsoft Phi-3 / Phi-4

Microsoft · 3.8B–14B

Allowed

MIT — commercial use allowed with no MAU cap; retain the license notice.

MIT LicenseOpen weights
Official license

Falcon 2 11B

TII · 11B

Allowed

Apache 2.0 — commercial use allowed, no MAU cap.

Apache 2.0Open weightsShip the license copy
Official license

IBM Granite 3.x

IBM · 2B–8B

Allowed

Apache 2.0 — enterprise-friendly, commercial use allowed, no MAU cap.

Apache 2.0Open weightsShip the license copy
Official license

AllenAI OLMo 2

Allen Institute for AI · 7B–13B

Allowed

Apache 2.0, fully open (weights + data + code) — commercial use allowed.

Apache 2.0Open weightsShip the license copy
Official license

Yi-1.5

01.AI · 6B–34B

Allowed

Apache 2.0 (Yi-1.5 relicensed) — commercial use allowed, no MAU cap.

Apache 2.0Open weightsShip the license copy
Official license

Llama 3.1 / 3.2 / 3.3 / Llama 4

Meta · 1B–405B

Conditional

Commercial use allowed under 700M monthly active users, with attribution, the "Llama" naming rule, and the Acceptable Use Policy.

Llama Community LicenseOpen weights≤ 700,000,000 MAUAttribution: Built with LlamaFine-tune naming ruleShip the license copyAcceptable-use policy
Official license

Gemma 2 / Gemma 3

Google · 2B–27B

Conditional

Commercial use allowed with no MAU cap, but the Gemma Prohibited Use Policy applies and must be passed through to your users.

Gemma Terms of UseOpen weightsShip the license copyAcceptable-use policy
Official license

Qwen2.5-72B

Alibaba Cloud · 72B

Conditional

Commercial use allowed below 100M monthly active users; above that you must request a license from Alibaba Cloud.

Tongyi Qianwen LicenseOpen weights≤ 100,000,000 MAUShip the license copyAcceptable-use policy
Official license

Cohere Command R / R+ (open weights)

Cohere · 35B · 104B

Not allowed

Open weights are CC-BY-NC 4.0 — non-commercial only. For a commercial product, use Cohere's API instead.

CC-BY-NC 4.0Open weightsAttribution: Cohere / C4AI attributionAcceptable-use policyNon-commercial weights
Official license

Mistral Large / Medium (open weights)

Mistral AI · 123B class

Not allowed

Open weights are research-only under the Mistral Research License. Commercial products need a Mistral Commercial License or the API.

Mistral Research LicenseOpen weightsAcceptable-use policyNon-commercial weights
Official license

OpenAI GPT-4o / GPT-5 / o-series

OpenAI · API only

Allowed via API

No weights to redistribute. Commercial use is allowed via the OpenAI API under the Business Terms; you own the outputs.

OpenAI Business TermsAPI onlyAcceptable-use policy
Official license

Anthropic Claude (Opus / Sonnet / Haiku)

Anthropic · API only

Allowed via API

No weights to redistribute. Commercial use is allowed via the Anthropic API under the Commercial Terms; you own the outputs.

Anthropic Commercial TermsAPI onlyAcceptable-use policy
Official license

Google Gemini (2.x API)

Google · API only

Allowed via API

No weights for the hosted models. Commercial use is allowed via the Gemini API under Google's API terms; you own the outputs.

Gemini API TermsAPI onlyAcceptable-use policy
Official license

Each row links its official license text. Verdicts are derived from the license category, not asserted. See the full source list and last-verified date below the tool.

How it works

This is a lookup-and-decision tool, not a calculator. Every model maps to a single canonical license — Apache 2.0, MIT, the Llama Community License, Google's Gemma Terms, the Tongyi Qianwen License, CC-BY-NC 4.0, the Mistral Research License, or a provider's API terms — and the commercial-use verdict is derivedfrom that license's category. Nothing is hand-asserted per row, so a model can never silently contradict its own license.

Each license sits in one of four buckets, and the verdict follows directly:

  • Permissive (Apache 2.0, MIT) → Allowed. Commercial use with no monthly-active-user cap; your only obligation is to keep the license / NOTICE file. Mistral 7B, Qwen3, DeepSeek, Phi, Granite and friends live here.
  • Conditional (Llama Community, Gemma, Tongyi Qianwen) → Conditional. Commercial use is allowed, but specific clauses apply: an MAU ceiling (700M for Llama, 100M for Qwen2.5-72B), an attribution string such as "Built with Llama", a fine-tune naming rule, and an acceptable-use policy.
  • Non-commercial (CC-BY-NC 4.0, Mistral Research License) → Not allowedfor the weights. Cohere Command R+ open weights and Mistral Large's research weights cannot ship in a paid product; the tool points you to the provider's API instead.
  • API terms-of-service (OpenAI, Anthropic, Gemini) → Allowed via API. There are no weights to redistribute; commercial use is governed by the provider's business terms and you generally own the outputs, subject to their usage policy.

The condition chips on each card — MAU ceiling, attribution, fine-tune naming, "ship the license copy", acceptable-use policy — are explicit per-model fields, each justified by a quoted clause in the data module, never inferred. To keep the catalogue honest, the build runs an internal cross-check: it re-derives every verdict a second way, from the license clauses alone, and asserts the two methods agree. The "18entries · cross-checked" badge means all of them reconcile. Filtering and search run entirely in your browser over this static dataset — no API calls, no keys, instant.

Worked examples

Startup fine-tunes Llama 3.3 70B for a 50,000-MAU SaaS

  1. Lookup → Llama family, license category = conditional.
  2. MAU check: 50,000 < 700,000,000 ceiling ⇒ commercial use Allowed.
  3. Conditions surfaced: display "Built with Llama", include the license copy,
  4. prefix the fine-tune name with "Llama", follow the Acceptable Use Policy.
  5. Verdict: ✅ ship it, with attribution.

Developer wants to bundle Cohere Command R+ weights in a paid app

  1. Lookup → Command R+ open weights = CC-BY-NC 4.0, category = non-commercial.
  2. NonCommercial clause ⇒ commercial use of the weights Not allowed.
  3. Tool recommends: switch to Qwen3 (Apache 2.0) or DeepSeek-V3 (MIT),
  4. or use Cohere's commercial API under their business terms.
  5. Verdict: ❌ not the weights — use the API or a permissive model.

Sanity check — DeepSeek-V3 weights for a commercial product

  1. Lookup → MIT License, category = permissive.
  2. No MAU cap, no attribution beyond the notice.
  3. Only obligation: retain the MIT copyright + license notice.
  4. Verdict: ✅ Allowed — the cleanest commercial option.

Frequently asked questions

Sources & references

Related tools

Rate this tool
Be the first to rate

Comments & feedback

Spotted a bug or want an improvement? Tell us — our team reviews every comment, and good ideas get built. Comments are public and anonymous.

Spotted a license that changed, or a model we should add?

Email me at [email protected] — most fixes ship within 24 hours.