Meta's Noninvasive Brain–Computer Interface Brain2Qwerty Achieves 61% Accuracy

2026-07-14 · Source: InfoQ · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Health & Medical Research · Depth: Advanced, quick

Summary

Meta has open-sourced Brain2Qwerty v2, a noninvasive Brain–Computer Interface (BCI) designed to decode sentences directly from human thoughts using electroencephalography (EEG) or magnetoencephalography (MEG) signals. This system achieved a 61% word accuracy rate, significantly outperforming other non-invasive methods which typically reach 8%. Brain2Qwerty v2 employs a three-stage deep-learning model, with MEG signals demonstrating superior performance (29% character error rate) compared to EEG (65%). The model also showed a 2.5x better character error rate than the baseline EEGNet. Meta has made both the model code and training data publicly available, aiming to advance research for individuals with communication impairments due to brain lesions. The project is part of Meta's broader Digital Brain initiative, which also includes tools like NeuralSet and NeuralBench.

Key takeaway

For research scientists developing noninvasive Brain–Computer Interfaces, Meta's Brain2Qwerty v2 demonstrates that 61% word accuracy is achievable through data scaling, not solely architectural breakthroughs. You should utilize the open-sourced model and training data to advance communication aids, focusing on MEG signals for improved decoding in neurological applications.

Key insights

Meta's Brain2Qwerty v2 significantly advances noninvasive BCI thought-to-text decoding, achieving 61% accuracy by scaling data.

Principles

• Non-invasive BCI accuracy scales log-linearly with data.
• MEG signals yield better decoding than EEG for BCI.
• Open-sourcing fosters neuroscience and BCI advancements.

Method

Brain2Qwerty v2 employs a three-stage deep-learning model: an Encoder for character prediction from brain signals, an Aligner for word grouping, and an LLM for final output generation.

In practice

• Brain2Qwerty v2 corrects "typographical" errors in decoded thoughts.
• Access Brain2Qwerty v2 code on GitHub and data on Huggingface.

Topics

• Brain-Computer Interface
• Noninvasive BCI
• Electroencephalography
• Magnetoencephalography
• Deep Learning Models
• Open-Source AI
• Digital Brain Project

Code references

• facebookresearch/brain2qwerty

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• Meta's brain-reading AI leaves letters behind — Meta's Brain2Qwerty v2 decodes full sentences from non-invasive brain scans, nearing surgical accuracy for communication.
• Text Without Typing — Brain2Qwerty v2 translates non-invasive brain activity into text, demonstrating improved performance with increased and multi-subject training data.
• Accurate Decoding of Natural Sentences from Non-Invasive Brain Recordings - AI at Meta — Non-invasive brain-to-text decoding can achieve high accuracy by scaling data and integrating advanced AI techniques.
• Meta's non-invasive brain-to-text AI is closing the gap with surgical implants — Brain2Qwerty v2 reconstructs sentences from non-invasive brain signals, achieving 39% word error rate by using more data and an asynchronous approach.
• Brain-to-Text Without Surgery — Non-invasive brain-to-text technology, exemplified by Brain2Qwerty v2, is advancing rapidly, enabling communication from brain signals without surgery.
• Is EEG-to-Text Feasible in Real-World Scenarios? An In-Depth Analysis Using a Neuropsychology-Inspired Benchmark — New research demonstrates teacher-forcing-free EEG-to-Text decoding is feasible by addressing EEG instability with a novel benchmark.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by InfoQ.