Model · openorcaxopenchat-preview2-13b
OpenOrcaxOpenChat-Preview2-13B
Developer: Open-Orca (AlignmentLab.ai); OpenChat contributed the training packing algorithm
Availability: available · checked 2026-09-24 · source
Raw record: /data/models/openorcaxopenchat-preview2-13b.json
Fields
- id
- openorcaxopenchat-preview2-13b
- identifiers
- huggingface
- Open-Orca/OpenOrcaxOpenChat-Preview2-13B
- developer
- Open-Orca (AlignmentLab.ai); OpenChat contributed the training packing algorithm
- release_date
- 2023-07-31partial · sourcenote: HF repo creation date; may predate or postdate public release.
- weights_status
- open
- availability
- available · checked 2026-09-24 · source
- license
- llama2partial · sourcenote: From HF card metadata (`license: llama2`); the Llama 2 Community License applies to the base. Confirm against the license text.
- architecture
- family
- decoder_only
- note
- family from config 'architectures'. Positional encoding left not_recorded, matching llama-2-13b.
- n_layers
- 40recorded · source
- hidden_size
- 5120recorded · source
- n_heads
- 40recorded · source
- vocab_size
- 32002recorded · source
- context_length
- 4096recorded · source
- positional_encoding
- nullnot_recorded
- training_data
- A curated filtered subset of most of the GPT-4 augmented completions in OpenOrca; packed with OpenChat's MultiPack sampler.recorded · sourcenote: Card: 'This second preview release is trained on a curated filtered subset of most of our GPT-4 augmented data.' The exact subset size is not stated.
- techniques
- primary_sources
- record_history
- date:2026-09-24 · change:ingested as candidate from HF (Open-Orca/OpenOrcaxOpenChat-Preview2-13B@49d50e4348c4f454302e997fc39841f035738604) · by:ingest_hf.py ·date:2026-09-24 · change:preparer: developer, license note, training data, 2 edges (fine_tuned_from llama-2-13b, trained_on openorca); sources: card · by:claude (preparer, Sonnet 5) ·date:2026-09-24 · change:reviewed and promoted from staging (2 edge(s) accepted) · by:Wilson Pruitt ·
Parents
Weights descend
- fine_tuned_from → Llama-2-13b-hf declared source Card: 'We have used our own OpenOrca dataset to fine-tune Llama2-13B using OpenChat packing.'
Training data
- trained_on → OpenOrca declared source Card: 'trained on a curated filtered subset of most of our GPT-4 augmented data' from the OpenOrca dataset (its GPT-4 half distills GPT-4 outputs; see the openorca record). The subset, not the whole set, was used.
Children
Weights descend
- ← merged_from OpenOrca-Platypus2-13B declared source Card: 'OpenOrca-Platypus2-13B is a merge of garage-bAInd/Platypus2-13B and Open-Orca/OpenOrcaxOpenChat-Preview2-13B.' The card does not state the merge method, weights or ratios. Parent 2 of 2. Trained by Open-Orca per the card. Preview2 was staged by the preparer so this edge can resolve.
Read in
No station on the reading path has touched this record yet.