OPEN-1B: the world's first fully auditable 1.6B-parameter transformer training run
benfielding · x · 2026-09-15
OPEN-1B claims to be the world's first fully auditable transformer training run, showing every step of AI training and inference can be logged and audited, providing proof of training data, recipe, biases, and weights.
- Specs: 24-layer decoder-only transformer, 1.61B parameters (1.08B non-embedding), trained on 400B tokens from four permissively licensed sources (DCLM-Baseline, FineWeb-Edu, The Stack v2, Proof-Pile-2).
- Audit mechanism: checkpoints and hashes are committed to an immutable ledger; anyone can re-execute any portion of the training at home to verify what happened.
- The authors say the auditable-training technique scales to any LLM size and any ML ops, and could train watchdog models to monitor other models.
- Framing: AI can be built in the open, verified by ordinary people, like open source always has been.
Related event: Gensyn releases open-1b, the first fully auditable open-source model(9 posts)→
More from Models
- Anthropic's server-side rule reignites the case for owning open models over renting AI — alex_verem · 2026-11-02
- Epoch AI: open-weight GLM 5.3 exploited the patch gap to hack 6 Firefox releases at ~$1,000 each — rohanpaul_ai · 2026-11-02
- SemiAnalysis: safety classifiers sandbag scores of Opus 5.5 and GPT-6 Astra — morqon · 2026-11-02
- Liquid AI's 1.7B LFM3 becomes first local model to nail live fixtures lookup on a MacBook — NielsRogge · 2026-11-02
- Claude upgrade-page traffic rebounded to 17.6M in September, near May's 18.6M peak — firstadopter · 2026-11-02
- Open-source GLiNER-PII scrubs 13 identifiers from a discharge note in 550ms on GPU — MaziyarPanahi · 2026-11-02