← All models

Tencent HY 2.0 Think

Published Publisher Tencenthunyuan-2.0-thinking-20251109

Overview

Tencent HY 2.0 Think is a reasoning-oriented Hunyuan 2.0 language model. Tencent announced its release on November 9, 2025. The announcement describes improvements over the TurboS foundation in complex instruction following, multi-turn and long-document understanding, coding, agent tasks, and reasoning. (cloud.tencent.com)

Capabilities and uses

The model is suited to document analysis, coding assistance, multi-turn assistants, and agent workflows. Tencent’s TokenHub model list identifies deep thinking, web search, and function calling as supported capabilities. (main.qcloudimg.com)

Limits and availability

Tencent’s TokenHub listing specifies a 192k-token context window, a maximum input of 128k tokens, and a maximum output of 64k tokens. Limits and features can depend on the platform. (main.qcloudimg.com)

Tencent announced that access to this model through the legacy Tencent Hunyuan large-model service would end at 00:00 Beijing time on June 22, 2026, and recommended migrating to hy3-preview on TokenHub. This notice concerns the listed legacy service and should not be taken as proof of availability on every other platform. (cloud.tencent.com)

Pricing

Tencent Cloud’s mainland China Agent Development Platform announced historical rates effective March 13, 2026: CNY 0.0053 per 1,000 input tokens and CNY 0.0212 per 1,000 output tokens. Subscription customers used purchased packages to offset usage in platform units. These are platform-specific announced rates, not a verified current offer or a price for other Tencent services. Pricing source: Tencent Cloud pricing announcement; checked November 17, 2026. (cloud.tencent.com)

Official sources