# A Latent-Space Language Model Scales to 8.9 Billion Parameters and Beats OLMo-3 on Reasoning

NCP-ArchPreview adds concept-level prediction on top of next-token training and reports a 5.99-point gain on the grade-school math benchmark GSM8K.

- Published: 2026-09-11T06:22:22.078Z
- Canonical: https://polylog.news/ai/2026-09-11/a-latent-space-language-model-scales-to-8-9-billion-paramete
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.CL](https://arxiv.org/abs/2609.10715), [arXiv cs.CL (abstract)](https://arxiv.org/abs/2609.10715)

Next-token prediction has been the pretraining objective for every major language model, and the persistent criticism is that it forces a model to commit to surface tokens before it has settled on meaning. NCP-ArchPreview tests an alternati…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-09-11/a-latent-space-language-model-scales-to-8-9-billion-paramete (subscription information: https://polylog.news/pricing).