User Feedback Thread: PiCo-1B
Hey everyone! 👋
I've been absolutely blown away by how many of you have been trying out PiCo-1B since it launched — thank you, truly. It means a lot to see this little 1B model finding its way into your projects.
I'd love to hear from you! If you've been using PiCo-1B — whether for prototyping, edge deployments, fine-tuning experiments, or just messing around — I'd really appreciate it if you could drop a comment below about your experience. Things like:
- 🎯 What you're using it for
- ⚡ How it's performing for your use case
- 💡 Any surprises (good or bad!)
- 🚀 Features or improvements you'd like to see
Your feedback directly helps shape where the model goes next, and it also helps other folks in the community decide if PiCo-1B is right for their project.
No pressure at all — even a quick line or two would make my day. Thanks again for being part of this journey!
Make the gear turn
— ARC Develop Team
📜 License & Usage Policy
Effective Date: June 25, 2026
We are pleased to release PiCo-1B under the PiCo-1B Open RAIL-M License. This license is designed to foster open and responsible AI development by balancing broad accessibility with essential safeguards against misuse.
Please carefully review the full license terms before downloading, using, or distributing the model. The complete legal agreement is available in the LICENSE file in this repository.
✅ What You Are Permitted to Do
This license grants you a worldwide, royalty-free, non-exclusive right to:
- Use the model for research, development, and commercial purposes.
- Modify and create Derivatives of the model.
- Distribute and sublicense the model or its derivatives, provided that all sublicenses include the same use-based restrictions.
- Integrate the model into commercial products, services, or hosted/API offerings.
⚠️ Use-Based Restrictions (Prohibited Uses)
To ensure the model is used ethically and responsibly, the following uses are strictly prohibited:
- Illegal or Unlawful Activities – Violating any applicable national or international laws or human rights treaties.
- Harmful Content Generation – Generating or promoting defamatory, obscene, violent, terrorist, or sexually explicit content (including CSAM), or content that promotes self-harm, suicide, or addiction.
- Discrimination & Hate Speech – Generating content that discriminates against, harasses, or disparages individuals or groups based on race, gender, religion, age, disability, or any other protected characteristic.
- Privacy Violations & Surveillance – Creating deepfakes or impersonations without explicit consent; unauthorized facial recognition, biometric tracking, or surveillance; spreading disinformation or engaging in fraud.
- Professional Advice Replacement – Using the model to replace professional medical, legal, or financial advice.
- Automated High-Stakes Decision-Making – Deploying the model in systems that significantly impact individuals' employment, housing, credit, or legal status without meaningful human oversight and appeal mechanisms.
📋 Your Obligations
By using PiCo-1B, you agree to:
- Retain the License – Include a complete copy of this license with any distribution of the model or its derivatives.
- Pass Along Restrictions – Ensure that any derivatives you create are subject to at least the same use-based restrictions outlined in Section IV of the license.
- Provide Attribution – Clearly attribute any public use or publication of model outputs to "PiCo-1B by ArcOffical", and state that the outputs are AI-generated and may contain errors.
- Respect Trademarks – Refrain from using the names "ArcOffical", "PiCo", or "PiCo-1B" to endorse or promote your products without prior written permission.
📄 Full Legal Text
The above is a non-binding summary for your convenience. The complete, legally binding terms and conditions are specified in the LICENSE file located in the root of this repository. We strongly encourage you to read it in its entirety.
📬 Contact & Inquiries
If you have any questions regarding this license or need to report a potential violation, please reach out to us at:
ArcOffical
Disclaimer: This summary is provided for informational purposes only and does not constitute legal advice. Please consult the full LICENSE file or a legal professional for interpretation of your rights and obligations.
the model is not working for me, using an q4_k_m format and it just types so random stuff, tried a lot of stuff to fix this, would like to help me
Hey! Thanks for flagging this. The reason it's outputting gibberish comes down to two main things:
First, PiCo-1B is a base model, meaning it's designed to auto-complete text rather than hold a conversation. If you run it in a standard 'Chat' interface, it panics and outputs random data. Try switching your software to Text Completion / Notebook mode instead.
Second, because 1B models are already so small, heavy compression like Q4_K_M can easily corrupt their logic. Try grabbing the Q8_0 version or the whole model instead—it's way more stable and shouldn't output random stuff. Also, make sure your local software (Ollama/LM Studio) is fully updated!
thanks, problem is solved, but my hardware is not that great and your model solved this, your model has stable t/s decode even with huge context window.
That's awesome to hear, and thank you so much for the kind words! 🙌
We're really glad PiCo-1B is running well on your setup — stable decode speeds with long context on modest hardware is exactly what we built it for. If you ever hit another snag or have ideas for improvements, feel free to drop another comment anytime. Happy building! 🚀
And also can u help me to press the like button next to the model name? That really helps!