PulseAugur
EN
LIVE 00:16:10

Tencent's Hy3 open MoE model beats GLM-5.1 in blind test

Tencent has released Hy3, a 21 billion active parameter open Mixture-of-Experts model that outperformed Zhipu's GLM-5.1 in a blind test. The model, which has 295 billion total parameters and a 256K context window, is available under the Apache 2.0 license. Hy3 is accessible via OpenRouter and Tencent Cloud, with a promotional period offering free inference until July 21, 2026. AI

IMPACT Sets a new benchmark for open-source MoE models, potentially influencing routing decisions and competition in the LLM landscape.

RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on Towards AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Tencent's Hy3 open MoE model beats GLM-5.1 in blind test

COVERAGE [1]

  1. Towards AI TIER_1 English(EN) · Kashif Mehmood ·

    Tencent Hy3 Review: 21B-Active Open MoE That Beat GLM-5.1 in a Blind Test

    <div class="medium-feed-item"><p class="medium-feed-image"><a href="https://pub.towardsai.net/tencent-hy3-review-21b-active-open-moe-that-beat-glm-5-1-in-a-blind-test-f2e8700d5a5d?source=rss----98111c9905da---4"><img src="https://cdn-images-1.medium.com/max/1376/1*uQUTgoDaALKK6nF…