Baidu Releases ERNIE 5.5 Weights; Flash Base Claims Same-Size SOTA Across Benchmarks
xiaohu · x · 2026-11-05
Baidu announced at its World Conference 2026 that ERNIE 5.5 model weights are now open, with ecosystem support from the vLLM and SGLang communities (GitHub, Hugging Face, and blog links provided).
ERNIE 5.5 Flash Base reportedly achieves same-size SOTA across knowledge, reasoning, math, code, multilingual, and long-context benchmarks, surpassing DeepSeek-V4-Flash-Base, Hy3-preview-Base, MiMo-V2.5-Base, and others. Baidu also open-sourced FlashMask, a mask-aware distributed attention implementation for long-context training; FlashMask-4 supports Hopper and Blackwell GPUs with runtime JIT-compiled CuTe DSL kernels.
More from Models
- Unsloth Ships Dynamic V3.1 Quants Claiming 10-15% Accuracy Gains, Reddit Skeptical — fiocalisti · 2026-11-12
- KRAFTON Open-Sources Raon-1 nano: 1.5B Effective Params, 262K Context, On-Device Ready — pmttyji · 2026-11-12
- Robot build test: GPT-6 Astra stabilizes two-board structure, Claude Fable 5.2 falls short — qinzytech · 2026-11-12
- Zvi dissects Anthropic's latest system card, disputing the autonomy threshold call — Don't Worry About the Vase (Zvi) · 2026-11-12
- antirez backs low-bit DeepSeek 4.1 Flash over Qwen 27B dense on Apple M5 Macs — antirez · 2026-11-12
- Gemini re-renders images as text when vision is removed, dropping only 6% WSR — zacharynado · 2026-11-12