Official release date: 2026-11-20, explicitly dated in the model card rather than inferred from repository creation. Text/image input and text output. 26.4B total, 3.4B active parameters; the middle 12 of 24 layers execute three times with shared weights, with 5.4B effective active parameters reported. Roughly 20GB mixed NVFP4/MXFP8 checkpoint. The card recommends 32GB or more for Apple silicon memory or other GPU VRAM; this is not an AraoAI hardware reproduction. Apache 2.0 is declared alongside an Acceptable Use Policy: https://thinkingmachines.ai/model-acceptable-use-policy/ . No independent AraoAI benchmark yet.
0
Quantized builds
0
Benchmark records
Quantized builds
Community and official quantized releases; "linked records" counts only benchmarks tied to that exact release.
No quantized builds cataloged yet.
Benchmark data
No published benchmark records yet — data is curated and sourced entry by entry by the operators.
Community discussion
No discussions yet — start the first one.
Start a discussion