# Shanghai AI Lab and SJTU's LUMIA Lab release 8.9B open-weight model NCP-ArchPreview

NCP-ArchPreview is released under Apache 2.0. The labs report that, after training on 5.73 trillion Dolma 3 tokens, it matches OLMo-3-7B’s final Stage 1 loss with 51.3% of the tokens, a 1.95× convergence gain.

Language: en
Time zone: America/Los_Angeles
Current time (UTC, cached as of): 2026-11-10T07:31:00.000Z
Query date (America/Los_Angeles): 2026-11-09

HTML: https://didcodexreset.com/news/118b22ddcb5a464590032f42.html

Content language: en
Localization state: sameLanguage

Source: ModelScope · Published 9/22/2026, 22:08:08

Shanghai AI Lab and SJTU’s LUMIA Lab have released NCP-ArchPreview, an open-weight 8.9B language model under the Apache 2.0 license. The weights are in the [NCP-ArchPreview collection](https://modelscope.ai/collections/Shanghai_AI_Laboratory/NCP_ArchPreview), and a [paper](https://modelscope.ai/papers/2609.10715) is also available.

The labs report that the model was trained on 5.73 trillion Dolma 3 tokens and reaches OLMo-3-7B’s final Stage 1 loss with 51.3% of the tokens, a 1.95× convergence gain. They report that its Stage 1 macro-average rises from 46.59 to 49.04, with gains of 5.99 on GSM8K and 4.28 on HumanEval.

NCP jointly predicts tokens and a concept sequence at one-quarter the length, then feeds those concepts back to guide generation. Domain adaptation updates only the 17 million-parameter concept module while the token backbone stays frozen. The labs report that concept-conditioned drafting improves mean accepted length by 4.17%, with negligible overhead.

Tags: NCP-ArchPreview, Open-weight models, Language models, Shanghai AI Lab, Apache 2.0

[View original post](https://x.com/ModelScope2022/status/2102624028582133927)