DEV.co
AI Frameworks · eugeneyan

open-llms

open-llms is a curated list of large language models licensed for commercial use, including Apache 2.0, MIT, and OpenRAIL-M models. It catalogs model parameters, context lengths, release dates, and direct links to checkpoints and papers for builders evaluating open alternatives.

Source: GitHub — github.com/eugeneyan/open-llms
12.8k
GitHub stars
977
Forks
Unknown
Primary language
Apache-2.0
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositoryeugeneyan/open-llms
Ownereugeneyan
Primary languageUnknown
LicenseApache-2.0 — OSI-approved
Stars12.8k
Forks977
Open issues13
Latest releaseUnknown
Last updated2025-02-13
Sourcehttps://github.com/eugeneyan/open-llms

What open-llms is

The repository indexes 20+ open LLMs ranging from 70M to 176B parameters with context windows from 512 to 84k tokens, sourced from providers like EleutherAI, Google, Stability AI, and Yandex. Models span text-to-text (T5), RNN-based (RWKV), and instruction-tuned architectures (Dolly, StableLM) with documented licensing compliance.

Quickstart

Get the open-llms source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/eugeneyan/open-llms.gitcd open-llms# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Commercial product evaluation and prototyping

Use this resource to identify and compare open LLMs with explicit commercial licenses (Apache 2.0, MIT, OpenRAIL-M) before committing to proprietary APIs. Rapidly prototype with vetted checkpoints from HuggingFace and GitHub.

On-premises or edge deployment

Filter by parameter count and context length to select models suitable for constrained environments (e.g., DLite 1.5B, Cerebras-GPT 1.3B) without licensing friction or data residency concerns.

Research and benchmarking

Access linked papers, model architectures (RNN vs. Transformer), and training methodologies across a timeline (2019–2023) to understand model evolution and inform architectural choices for custom fine-tuning.

Implementation considerations

  • Verify each model's actual license file before deployment; some (e.g., ChatGLM, StableLM CC BY-SA-4.0) have compliance or attribution obligations that differ from stated summaries.
  • Context length and parameter size directly impact hardware requirements; benchmark inference latency and memory footprint for your target hardware (CPU, GPU, edge device).
  • Models range from 2019–2023; evaluate whether model age aligns with your domain (older models may lack knowledge cutoff or safety tuning).
  • Instruction-tuned variants (Dolly, MPT-7B-Instruct, h2oGPT) may perform better for chat/task workflows than base models; confirm downstream accuracy for your use case.
  • Checkpoint locations span multiple platforms (GitHub, HuggingFace, Cerebras); plan for multi-source artifact management and fallback retrieval.

When to avoid it — and what to weigh

  • You need production-grade SLAs and support — This is a static curated list, not a managed service. No guarantees on model performance, security patches, or availability. Each model has its own maintenance status (Unknown for most).
  • You require license certainty for restrictive jurisdictions — ChatGLM lists a custom license with 'usage restrictions' and 'may require registration.' Bloom uses OpenRAIL-M which may have regional compliance complexity. Requires legal review for high-risk deployments.
  • You need unified version management and rollback — This lists disparate checkpoints across multiple hosting providers (HuggingFace, GitHub, EleutherAI). Integration, versioning, and rollback strategies are your responsibility.
  • You expect real-time model updates or feature parity with commercial LLMs — Latest release is 'n/a'; last pushed 2025-02-13 indicates curation pauses. No roadmap or versioning for individual models. Use as a reference, not a live service.

License & commercial use

The list predominantly features Apache 2.0 (permissive OSI-approved) and MIT models. Bloom uses OpenRAIL-M v1 (open but with responsible AI guardrails). ChatGLM lists a custom license with usage restrictions and potential registration requirements. StableLM uses CC BY-SA-4.0 (copyleft with attribution). Legal review recommended for Bloom and ChatGLM in regulated industries.

Most models explicitly support commercial use; however, this is NOT a license grant. Each model's commercial viability depends on its stated license and your jurisdiction. Apache 2.0 and MIT models are generally safe for commercial deployment with attribution. Bloom (OpenRAIL-M) and ChatGLM require additional due diligence. Verify directly with model authors before production use; this list is informational only.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityNeeds review
Deployment complexityHigh
DEV.co fitGood
Assessment confidenceHigh
Security considerations

No security analysis provided. Each model carries inherited risks from its training data and architecture; review model cards and threat models independently. Inference-time poisoning or prompt injection vulnerabilities are not addressed. Plan for isolated inference environments if handling sensitive data. No disclosure process or security contact listed for the repository itself.

Alternatives to consider

Hugging Face Model Hub + Model Cards

Hosts the same models with built-in filtering, versioning, and community-contributed model cards including safety disclaimers. Offers centralized discovery but lacks curated licensing summary.

Together AI, Replicate, or other LLM-as-a-Service platforms

Provide managed inference, SLAs, and support for many of these same models (Cerebras-GPT, StableLM, Dolly). Trade open-source flexibility for operational simplicity if no on-premises requirement.

Custom fine-tuning on proprietary models (OpenAI API, Anthropic)

Eliminate licensing uncertainty and gain vendor support, but sacrifice data sovereignty, cost predictability, and on-premises deployment options.

Software development agency

Build on open-llms with DEV.co software developers

Devco's AI development team helps you select, integrate, and optimize open models for production. We handle licensing review, infrastructure setup, and fine-tuning. Let's discuss your requirements.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

open-llms FAQ

Can I use these models in a commercial product without paying royalties?
Most models listed (Apache 2.0, MIT) permit commercial use; however, review the specific license file for each model. Some (ChatGLM, Bloom) have restrictions or require attribution. Consult legal counsel before shipping.
What is the largest model I can deploy on consumer hardware?
Models up to ~7B parameters (Dolly-7B, StableLM-7B, RedPajama-INCITE-7B) can run on a high-end consumer GPU (24–40GB VRAM) with quantization. Larger models (Bloom 176B, YaLM 100B) require multi-GPU or cloud infrastructure.
How do I update a model if a newer version is released?
This list is a snapshot. Monitor individual model repositories (EleutherAI GitHub, Stability AI releases, HuggingFace org pages) for updates. You must manage versioning and rollback in your own deployment pipeline.
Are these models production-ready?
Models vary. Instruction-tuned variants (Dolly, h2oGPT, StableLM) are closer to production than base models. Evaluate output quality, hallucination rates, and latency for your specific use case. No SLA or support implied by this list.

Software developers & web developers for hire

Need help beyond evaluating open-llms? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and ai frameworks integrations — and maintain them long-term.

Ready to evaluate and deploy an open LLM?

Devco's AI development team helps you select, integrate, and optimize open models for production. We handle licensing review, infrastructure setup, and fine-tuning. Let's discuss your requirements.