🎯 Quantization-aware training

The items that name it 7

• Beyond 4 bits: Data-Centric Post-training Quantization and Recovery
  paper2026-10-20It is written here as Quantization-aware training (QAT).

  QAT

• [AINews] not much happened today
  article2026-10-20

  Used for Nemotron 3.5 Bolt down to 2-bit so it fits in 24GB.

• AnyLanguageModel 1.0: One Year of Local Models on Apple Platforms
  article2026-10-20
• [AINews] High Taste Testers Ep. 1
  article2026-10-17It is written here as Quantization-aware training (QAT).

  Nemotron 3.5 Bolt uses native 2-bit QAT to fit a 74B model in about 24GB.

• [AINews] High Taste Testers Ep. 1
  article2026-10-17It is written here as Quantization-aware training (2-bit).

  2-bit

• Sub-1-Bit LLM Compression via Latent Factorization
  repo2026-10-08It is written here as Quantization-Aware Training (QAT).

  QAT

• Quantization-Aware Healing: a compressed, 4-bit model that outperforms its full-precision original
  article2026-08-25It is written here as Quantization-Aware Training (QAT).

  QAT