Grok 4.5 Auto-Optimizes Speech-to-Text
Nilofer_tweets · x · 2026-07-15
Someone ran an automated research loop using Grok 4.5 to make Parakeet TDT 0.6B v3 run faster on CPUs.
The results:
- Identified that the encoder takes up about 98% of the runtime, so experiments focused there.
- Switching the encoder from the standard Hub dynamic INT8 path to static quant yielded significant performance gains.
- On an AMD EPYC machine with 8 vCPUs, throughput increased from roughly 26x real-time to 54x real-time, a 2.1x speedup.
- On Mac, long audio scenarios also saw a 1.4x speedup.
More from coding & agent
- Gergely Orosz: Shipping 10x PRs With AI Agents, Sites Fill With Small Regressions — ducha_aiki · 2026-09-11
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Astra storyboards plus Minimax H3 per-shot generation boost video success rates — Hailuo_AI · 2026-09-11
- Codex tip: use Sol with Astra and Luna sub-agents to save usage — pvncher · 2026-09-11
- agents-best-practices: a provider-neutral Agent Skill for designing and auditing agentic harnesses — tom_doerr · 2026-09-11
- Cognition's SWE-2 uses a KKT duality argument in RL to shift the effort Pareto curve — YouJiacheng · 2026-09-11