Brain2Qwerty

Overview

Brain2Qwerty is Meta FAIR’s non-invasive brain-to-text research pipeline that decodes typed sentences from MEG (magnetoencephalography) signals without requiring surgical implants. The v2 release (July 2026) hits ~39% average word error rate (61% accuracy) on typed sentences with the best participant at 22% WER (78% accuracy). Surgical implants still sit below 2% WER, so the gap remains real, but the non-invasive number is a meaningful research milestone for a modality that requires no surgery.

Timeline

• 2026-07-02-AI-Digest — Meta FAIR releases Brain2Qwerty v2 — a non-invasive MEG-signal-to-text pipeline hitting ~39% average word error rate (61% accuracy) on typed sentences, with the best participant at 22% WER (78% accuracy). Surgical implants still sit below 2% WER; the gap is real, but the non-invasive number is a meaningful research milestone. A research release, not a product. Meta’s public-lab BCI work continues to surface as a “quietly serious” thread inside the broader Meta AI narrative, distinct from the wearables and open-weights stories.

Key Developments

1. Non-Invasive Brain-to-Text Milestone (July 2026): ~39% average WER / 61% accuracy on typed sentences via MEG signals, best participant at 22% WER (78% accuracy). Surgical implants still sit below 2% WER, so parity is not the read — the read is that a non-invasive modality (no surgery) is now producing usable numbers for a research pipeline.

Related

See also: Meta, MOC - Major Companies.