Release History • Verifiable Benchmarks

Product Releases & Performance Benchmarks

Every update documented with verified latency metrics, memory footprints, and architectural milestones.

Verified Performance Bake-off (October 2026 Test Suite)

Evaluated against real-world dictation history: 100 test dictations, 60 minutes of audio, 6,670 words tested on Apple Silicon M-Series.

Metric Wispr Flow (Cloud) Speakeasy (On-Device Core ML) Advantage
Median Latency 478 ms (Cloud gRPC roundtrip) 431 ms (Local Core ML) 47 ms Faster
Speech Error Rate vs Ground Truth Baseline (4.91%) 4.84% WER Higher Accuracy
Speaker Diarization Accuracy Cloud API Dependent 4 of 4 Known-Answer Clips Verified 100% Local
Disk & Model Footprint ~950 MB (Electron + Webpack) 423 MB (Tauri + Rust + Weights) 55% Smaller
Runtime Memory ~1,200 MB ~480 MB 60% Less RAM

v1.0.0 — Production Core ML Release

2026-10-27

Complete clean-room implementation built in Rust and Apple Core ML. All 23 milestone specifications completed and verified.

Q1 2027 Roadmap — Cross-Platform Expansion