News tagged "moe"

Editorial diagram showing three routing lanes for Tencent Hy3 - fast, slow, and hybrid reasoning modes - with cost and latency tradeoffs on a clean muted-color policy decision map

Tencent Hy3 Lands on TokenHub: What the 295B MoE Model's Three Reasoning Modes Mean for Your API Routing Policy

Tencent officially launched Hy3 on July 6, setting live API access on TokenHub and rolling out to multi-provider AI gateways globally. The Tencent Hy3 API routing operator decision now involves three reasoning modes and a price point below $0.15 per million input tokens.

July 7, 2026via Tencent

reinforcement-learning

GSPO: Stable MoE RL Training Without Routing Replay

Qwen's GSPO shifts optimization from token-level to sequence-level, eliminating Routing Replay overhead and enabling stable large-scale RLHF training for Qwen3 models.

May 19, 2026via Qwen

Help & contact