
Hy4 flash launches with lower API prices and 1M context
Tencent has introduced Hy4 flash, a 224B-parameter multimodal model with 12B active parameters per token, a 1M-token context window and lower API pricing than Hy4.
The company says its hybrid architecture, self-distillation, reinforcement learning and 4-bit inference stack support decoding speeds above 150 tokens per second.
Tencent reports that Hy4 flash slightly exceeds Hy4 preview on overall task performance, and it is available across its services and as open weights on several code and model platforms.
No comments so far, maybe you want to be first?
Gu



