Tagged articles

MoE architecture

5 articles · Page 1 of 1

Aug 4, 2026 · Artificial Intelligence

Tencent Hunyuan Hy ASR 3.0 Preview Raises Accuracy, Dialect Support, and Robustness

On August 4, Tencent released the Hy ASR 3.0 preview, a speech‑recognition model built on the Hy3 large‑language model that combines a MoE architecture, massive unsupervised audio training and multi‑stage reinforcement learning to cut word‑error rates to around 3 % across Mandarin, English and Cantonese, while improving dialect coverage, context understanding and stability in noisy environments, and is now offered as a cloud API.

ASRHy ASR 3.0Large Language Model

0 likes · 6 min read

Tencent Hunyuan Hy ASR 3.0 Preview Raises Accuracy, Dialect Support, and Robustness

SuanNi

Jun 12, 2026 · Artificial Intelligence

Kimi K2.7 Code Goes Open: 30% Token Savings and Major Coding Performance Boost

Kimi K2.7 Code, now open‑source on HuggingFace, reduces token consumption by ~30% and boosts coding benchmark scores—Kimi Code Bench v2 climbs from 50.9 to 62.0, Program‑Bench from 48.3 to 53.6, MLS Bench Lite from 26.7 to 35.1—narrowing the gap with GPT‑5.5 and Claude Opus, all built on a 1‑trillion‑parameter MoE architecture with INT4 quantization and a 256K‑token context.

Kimi K2.7LLM benchmarksMoE architecture

0 likes · 6 min read

Kimi K2.7 Code Goes Open: 30% Token Savings and Major Coding Performance Boost

Machine Heart

Apr 3, 2026 · Artificial Intelligence

Google Open‑Sources Gemma 4, Outperforming a 13×‑Larger Qwen 3.5

Google DeepMind released the open‑source Gemma 4 family—four model sizes ranging from 2 B to 31 B parameters, supporting text, images, video and audio, with up to 256 k token context, Apache 2.0 licensing, and benchmark results that place it on par with the 397 B Qwen 3.5 despite being far smaller.

Apache 2.0Gemma 4Google DeepMind

0 likes · 11 min read

Google Open‑Sources Gemma 4, Outperforming a 13×‑Larger Qwen 3.5

JavaEdge

Jul 28, 2025 · Artificial Intelligence

Why Kimi K2 Is the Next Open-Source LLM Challenging DeepSeek

The article examines Kimi K2, Moonshot AI’s open‑source large language model, detailing its MoE architecture, low‑cost pricing, agentic capabilities, performance comparisons with Claude and DeepSeek, and real‑world developer experiences, while discussing its potential impact on the AI landscape.

AI CostKimi K2MoE architecture

0 likes · 8 min read

Why Kimi K2 Is the Next Open-Source LLM Challenging DeepSeek

Architects' Tech Alliance

Feb 18, 2025 · Artificial Intelligence

How DeepSeek’s Latest Models Redefine AI Performance and Industry Adoption

The DeepSeek report details rapid model releases from 2024 onward, highlighting innovations such as model distillation, a 671 B MoE architecture, FP8 mixed‑precision, and the Janus‑Pro multimodal framework, while also documenting major cloud and chip providers' integration of these models into their services.

AI industry adoptionDeepSeekLarge Language Models

0 likes · 10 min read

How DeepSeek’s Latest Models Redefine AI Performance and Industry Adoption