# miniTE trains learner updates 1.33–1.37× faster than BF16 on 8×B300

> Original title: 🚀 miniTE: 1.33–1.37× BF16 speed for learner updates on 8×B300.

- Source: LMSYS Org
- Published: 2026-10-27T18:33:17.000Z
- Bad Influence: https://badinfluence.news/items/gier7teecklb85dqgdl0pfq1c
- Original: https://x.com/lmsysorg/status/2115149852720152642

## Summary

InfiX.ai's miniTE, a low-precision training framework, runs learner updates 1.33–1.37× faster than BF16 on 8×B300 GPUs, according to LMSYS Org. The framework diagnoses training–sampling mismatch from quantization to token probabilities and uses TRIAGE to control risky updates. It also reduces compute overhead with grouped GEMM, contiguous buffers, and fusion, and speeds up NVFP4 computation.

## Post

🚀 miniTE: 1.33–1.37× BF16 speed for learner updates on 8×B300.

InfiX.‌ai’s miniTE is a low-precision training framework that reduces training–sampling mismatch and speeds up NVFP4 computation.

Highlights:
- Diagnose mismatch from quantization to token probabilities.
- Control risky updates with TRIAGE.
- Reduce compute overhead with grouped GEMM, contiguous buffers, and fusion.

Read the full report: http://lmsys.org/blog/2026-10-25-minite

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHVqFj6TaEAAcnIp.jpg&mode=thumb&exp=1793404800&sig=ebd544d7acdc64c6)