Tencent Hy-MT2-Pro

Tencent Hy-MT2-Pro sits inside Tencent's Hy family as a closed-weights, text-only reasoning model that aggregator listings describe as targeting coding, instruction following, and agent tasks. Because the model is marked closed rather than open-weights, the underlying architecture, training data, and parameter counts are not disclosed publicly, and the available evidence is limited to a single third-party pricing page rather than any Tencent technical report or model card. In practice this means the model's design lineage cannot be drawn from peer-reviewed or vendor-published material; only its family affiliation and intended use cases are explicitly grounded in the supplied evidence.

For practical fit, the model is positioned for developer-oriented workloads where focused reasoning over short-to-medium contexts matters more than broad general knowledge. Its listed operational envelope of an 8,000-token context window and a 4,000-token output ceiling makes it better suited to contained coding sessions, structured tool-mediated agent loops, and instruction-following tasks than to long-document reasoning or extended multi-turn synthesis. Teams looking for a compact, task-specialized reasoning model within the Hy ecosystem can consider it for these scoped agent and coding scenarios, while heavier long-context work would push past its published limits.