Baidu's ERNIE Evolve boosts ERNIE 5.5 Flash generation speed 6.2x on Kunlunxin P800
BaiduResearch · x · 2026-11-10
Baidu Research detailed three inference optimizations ERNIE Evolve applied to ERNIE 5.5 Flash, with full optimization-path curves: low-latency inference on Kunlunxin P800 reaching 623.6% of baseline generation speed via 15 steps (speculative decoding, fused mHC, native attention, cached expert execution, etc.); deterministic inference at 133.7% of baseline throughput while preserving bitwise determinism; and MoE scheduling/kernel improvements pushing decode speed to 114.9% with no accuracy loss. All optimizations passed rigorous load testing and correctness validation.
Related event: Baidu Releases Open-Source ERNIE Evolve Automated Research System(6 posts)→
More from Infra
- simdjson disables 256-bit LASX on Loongson 3A6000: 128-bit LSX is 6.7% faster — lemire · 2026-11-10
- Vera launches gyms, benchmarks and agents for cloud infrastructure operations — tianyin_xu · 2026-11-10
- NVIDIA paper: masking ~10 'intrusion heads' recovers 57% of distractor accuracy loss — rohanpaul_ai · 2026-11-10
- Memory pricing itself out of demand: PC BOM share hits 35-38%, shipments plunge — tengyanAI · 2026-11-10
- Two engineers burned $500K in API usage in one month — $6M annualized — bingxu_ · 2026-11-10
- Flide: exact state forking speeds decision-model agents up to 9.5x — NationalUniversityofSingapore · 2026-11-10