Biological models

318 model records in release 2026-10-23-5347695500ed (release manifest, gzip). Showing 145–168; page 7 of 14.

Evaluated configurations, methods and pipelines are listed separately. Names alone do not establish equivalent models or checkpoints.

• hgT5

  DNA and genomes

  T5 model pretrained on the human genome by the GUANinE authors (Robson and Ioannidis 2023).

• HyenaDNA

  DNA and genomes

  DNA language model pretrained on the human reference genome with context lengths of 32 kb to 1 Mb.

• ICEBERG

  Metabolomics

  ICEBERG forward spectrum simulation model, used for retrieval by scoring candidates against simulated spectra.

• JanusDNA

  DNA and genomes

  Hybrid Mamba, attention and mixture-of-experts DNA model that combines autoregressive and masked pretraining to read sequences of up to 1 Mb in both directions at single-nucleotide resolution.

• JESTR

  Metabolomics

  Contrastive model that matches MS/MS spectra to candidate molecules for metabolite retrieval, pretrained with batch InfoNCE and fine-tuned with candidate InfoNCE.

• Jores plant promoter CNN

  DNA and genomes

  Convolutional neural network that predicts plant core promoter strength, measured by STARR-seq in tobacco leaves or maize protoplasts, directly from promoter sequence.

• Kimi-K2.6

  Cells and tissues

  General-purpose large language model, release K2.6 of the Kimi K2 series.

• LAMAR

  RNA and transcriptomes

  LAMAR RNA foundation model pretrained on mammalian genome and transcriptome sequences (2k and 4k context checkpoints).

• LangCell

  Cells and tissues

  Single-cell foundation model pretrained on cell-text pairs, initialised from Geneformer.

• LCDD-team

  Molecular interactions

  LCDD-team: three-stream GPS graph transformer affinity model from the CASP16 LCDD team.

• Llama 2 13B Chat

  Cells and tissues

  Chat-tuned 13-billion-parameter model in Meta's Llama 2 family of open-weight language models.

• Llama 2 70B Chat

  Cells and tissues

  Chat-tuned 70-billion-parameter model in Meta's Llama 2 family of open-weight language models.

• Llama 2 7B Chat

  Cells and tissues

  Chat-tuned 7-billion-parameter model in Meta's Llama 2 family of open-weight language models.

• Llama 3 70B Instruct

  Cells and tissues

  Instruction-tuned 70-billion-parameter model in Meta's Llama 3 family of open-weight language models.

• Llama 3 8B Instruct

  Cells and tissues

  Instruction-tuned 8-billion-parameter model in Meta's Llama 3 family of open-weight language models.

• Llama 3.1 70B

  DNA and genomes

  70-billion-parameter model in Meta's Llama 3.1 collection of multilingual auto-regressive Transformer language models, with grouped-query attention and a 128k-token context.

• Llama 3.2 1B Instruct

  Cells and tissues

  Instruction-tuned 1-billion-parameter text model in Meta's Llama 3.2 collection of open-weight language models.

• Llama 3.2 3B Instruct

  Cells and tissues

  Instruction-tuned 3-billion-parameter text model in Meta's Llama 3.2 collection of open-weight language models.

• Llama 3.3 70B Instruct

  Cells and tissues

  Instruction-tuned 70-billion-parameter text model in Meta's Llama 3.3 release of open-weight language models.

• Llama-3.1-8B

  Cells and tissues

  Language model in Meta's Llama 3 herd, which natively supports multilinguality, coding, reasoning and tool use.

• LlaSMol-Mistral-7B

  Protein function

  LlaSMol-Mistral-7B chemistry large language model, prompted for protein property questions.

• LLF

  Molecular interactions

  LLF: sequence-based protein-ligand affinity model with multi-stream GCN and CNN branches that keep local features.

• LOL-EVE

  DNA and genomes

  Language Of Life for Evolutionary Variant Effects: a conditional autoregressive transformer (CTRL-based, 12 layers) trained on 13.6 million mammalian promoter sequences, with gene, species and clade control codes, for zero-shot prediction of promoter indel effects.

• LucaOne

  DNA and genomes

  Transformer encoder pretrained jointly on nucleic acid and protein sequences from about 170,000 species, giving one representation for DNA, RNA and protein inputs.

This index reflects a dated catalogue, not an exhaustive census. Source checking does not mean independent reproduction; compare results only under compatible protocols, datasets and metrics.