Amazon Launches Nova Forge For Custom Models
Amazon unveiled Nova Forge at AWS re:Invent in Las Vegas, a service that lets companies train custom models using Nova checkpoints at pre-, mid-, and post-training stages. The company says Forge enables early proprietary-data injection to avoid capability regression and preserve base model strengths, and highlighted Reddit training a safety model on 23 years of moderation data. Amazon framed the move as prioritizing control and specialization over leaderboard benchmarks.
Key Points
- 1Introduces Nova Forge: access to Nova checkpoints at pre-, mid-, and post-training stages.
- 2Enables early proprietary data injection to preserve capabilities and avoid capability regression during customization.
- 3Lets companies retain weights, control safety, and build domain-expert models without full retraining costs.
Scoring Rationale
Official AWS product with strong enterprise applicability, but real-world impact depends on developer adoption and demonstrated results.
Sources
Public references used for this report.
Practice with real Retail & eCommerce data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Retail & eCommerce problems
