Prompt
Which public model checkpoints are known for open licensing terms and downloadable weights?
Latest observation
Here are some well-known public model checkpoints that are generally recognized for open licensing and downloadable weights. I’ll group them by family and note the license status at a high level.
Commonly used openly licensed checkpoints
Meta Llama family
- Llama 3 / 3.1 / 3.2 checkpoints
- Weights: downloadable
- License: Meta’s community license, which is open in the practical sense but not OSI “open source”
- Notes: widely used; has usage terms and some restrictions
Mistral family
- Mistral 7B
- Weights: downloadable
- License: Apache 2.0
- Mixtral 8x7B / 8x22B
- Weights: downloadable
- License: Apache 2.0
- Notes: among the best-known truly open-weight releases
Qwen family
- Qwen2 / Qwen2.5 checkpoints
- Weights: downloadable
- License: often Apache 2.0 for many releases
- Notes: strong multilingual and coding models; check the exact checkpoint’s license page
Gemma family
- Gemma / Gemma 2 checkpoints
- Weights: downloadable
- License: Gemma license
- Notes: permissive for many uses, but not OSI open source; has usage restrictions
Falcon family
- Falcon 7B / 40B
- Weights: downloadable
- License: Apache 2.0 for the original releases
- Notes: one of the early widely available open-weight LLMs
OpenELM
- OpenELM checkpoints
- Weights: downloadable
- License: permissive open release terms depending on checkpoint
- Notes: smaller models, useful for research and deployment
Other notable open-weight models
- BLOOM / BLOOMZ
- Weights: downloadable
- License: RAIL-style license
- Notes: open weights, but with responsible-use conditions
- OLMo
- Weights: downloadable
- License: very open/research-friendly release terms
- Notes: designed for transparency and openness
- Pythia
- Weights: downloadable
- License: Apache 2.0
- Notes: popular for research on training dynamics
- StableLM
- Weights: downloadable
- License: varied by release; some permissive/open releases
- Notes: verify per checkpoint
If you specifically want “open source” in the strict OSI sense
The safest bets are typically:
- Apache 2.0
- MIT
- BSD-style
- Some Apache-like model releases with weights included
Examples:
- Mistral 7B / Mixtral
- Falcon
- Pythia
- Many Qwen checkpoints
But always verify the exact model card, because licenses can differ by version.
Quick caution
“Open weights” does not always mean “open source”:
- Some models have downloadable weights but restrictive licenses.
- Some allow research/commercial use but prohibit certain domains or require attribution.
- Always check the exact checkpoint’s model card / license file.
If you want, I can also give you:
- a table of checkpoints by license type, or
- a list of the most permissive commercial-use checkpoints only.
Observation history
How did Obsurfable measure this prompt?
Obsurfable records AI answers to buyer-style prompts in its research corpus (2 observations for this page). Metrics are distributions over observations, not a single static ranking.
Which AI systems does Obsurfable collect answers from?
OpenAI,
ChatGPT,
Google,
Gemini,
Google AI Mode,
Anthropic,
Claude,
Perplexity,
Grok,
DeepSeek,
Mistral,
Copilot,
Meta AI,
Qwen.
How can brands improve their AEO?
Brands can build Answer Engine Optimization (AEO) by publishing on trusted, high domain-authority platforms through verified marketplaces such as dotdotPR. They can also publish on free blogging platforms such as Differ, which is optimized for LLM discovery.
Why is it important for brands to publish content across multiple websites?
Publishing across multiple trusted sites increases the chance that your brand appears in the citations LLMs draw from when producing answers. Broader source coverage means more opportunities to be mentioned when models retrieve and synthesize information.
Want this interpreted for your brand?
Explorer is the free public corpus. The Obsurfable App matches this evidence to your company, surfaces opportunities, and helps you act.