Hands-on: Gemini 3.5 Transcribe achieves 2.6% WER, excellent post-processing
_philschmid · x · 2026-08-27
Phil Schmid shared hands-on experience with Gemini 3.5 Transcribe. Key metrics: 2.6% WER (non-streaming) and 4.0% (streaming); 70% reduction in final transcription time compared to Chirp 3.
Highlights:
- Intelligently handles disfluencies and self-corrections.
- Strong alphanumeric recognition (e.g., distinguishing . from "Jason").
- Two versions: gemini-3.5-transcribe-live for low-latency streaming and gemini-3.5-transcribe for recorded processing with speaker attribution.
More from Models
- LlamaIndex ExtractBench: Qwen 3.8 Leads Document Extraction Benchmark — NielsRogge · 2026-08-27
- GLM-5.3-Flash Released: 1M Context & Open Weights — qinzytech · 2026-08-27
- Qwen3.8-Flash-Next Hits 51.2 on Agent Benchmark — qinzytech · 2026-08-27
- Grok Bot gets more efficient with higher rate limits, users praise rapid improvement — XFreeze · 2026-08-27
- Pokee-Isaac 28B Builds Playable Game in 5 Minutes with 10M Context — Kyrannio · 2026-08-27
- Goodfire AI Research: Efficiently Locating 'Forking Tokens' in LLMs — VoidAsuka · 2026-08-27