Interconnects AI · 2026-08-02 · notable
Nathan Lambert: 'Open artifacts #23' — open-model consolidation isn't happening
Nathan Lambert's Interconnects #23 argues consolidation predicted for 2026 isn't happening. Instead more labs — Thinking Machines' Inkling, Tencent's Hy3, Poolside's Laguna S 2.1 — keep training strong open models.
Interconnects #23 argues the open-model field is widening, not consolidating — more labs are training strong open models than predicted.
What is it?
Nathan Lambert and Florian Brand's Aug 2 Interconnects essay catalogs about ten open-model releases from the past month and pushes back on the consolidation thesis. The read: labs are treating token production as a business, and open-model diversity is growing, not shrinking.
How does it work?
Interconnects #23 walks through each recent release — Thinking Machines' Inkling (975B-A41B multimodal MoE), Tencent's Hy3 (295B-A21B, Apache 2.0), Poolside's Laguna S 2.1 (118B-A8B, fits on a DGX Spark), Meituan's LongCat-2.0 (1.6T on Ascend 910s), AMD's Instella-MoE (16B-A3B on Instinct cards), Swiss AI's Apertus v1.5 70B, and Motif-3-Beta (314B-A13B) — showing labs spread across the U.S., China, Europe, and specialty accelerators.
Why does it matter?
The essay pushes back on the 'consolidation is inevitable' storyline that has framed AI in 2025-26. If Nathan Lambert is right, open-model diversification is the actual pattern — more labs entering the field, more geographies training frontier open weights, and the Pareto frontier widening rather than compressing.
Who is it for?
open-source AI followers, ML researchers, infra leads