Understand the technology
Filter the public specialist and research weights in our registries. Compare the sources before choosing an artifact for your own trial.
Models to explore: 9
ML engineering
Valid held-out ML-engineering and external MLE-bench evaluations were not completed for 1.0. Plan your own bounded trial.
ML engineering
Valid held-out ML-engineering and external MLE-bench evaluations were not completed for 1.0. Plan your own bounded trial.
ML engineering
Valid held-out ML-engineering and external MLE-bench evaluations were not completed for 1.0. Plan your own bounded trial.
Defensive cybersecurity
Intended for human-reviewed defensive workflows. Review evaluation scope and failures before using a proposed change.
Defensive cybersecurity
Intended for human-reviewed defensive workflows. Review evaluation scope and failures before using a proposed change.
Defensive cybersecurity
Intended for human-reviewed defensive workflows. Review evaluation scope and failures before using a proposed change.
Experimental research
Experimental — not recommended for production
A character-training research artifact with documented trade-offs. It is not recommended for production.
Everyday questions and writing
Capabilities and trade-offs vary. Read the evaluation and limitations for this exact release.
Everyday questions and writing
Capabilities and trade-offs vary. Read the evaluation and limitations for this exact release.
Read the licence, check the runtime, and validate the exact artifact on your own tasks. A quantized file is a different artifact from the merged weights.
Read the model-file guide