The repository identifies Carbon-X as a fine-tune of Qwen/Qwen3-14B and is not gated.
The model is distributed under the Apache-2.0 licence. Its page provides Transformers examples using pipeline and AutoTokenizer/AutoModelForCausalLM, and lists compatibility with Hugging Face Inference Endpoints. The repository reports 14,822,046,720 BF16 parameters and a total weight-file size of 29,644,181,703 bytes, or about 29.6 GB.
The model page was created on October 21, 2026 and last modified on October 22, 2026. It is classified as a Transformers text-generation model and carries tags for Qwen3, DNA, genomics, biology, and conversational use.
The source does not provide a benchmark table, training-data description, evaluation protocol, or biological validation results. It therefore does not establish that Carbon-X outperforms larger models, nor that generated outputs are suitable for scientific or clinical use.