Tencent Releases Hy4 Flash Model at Lower Prices
The Flash-tier model lists at lower per-token prices than Tencent's flagship Hy4 and scores slightly lower on tested tasks.
Updated on Nov. 16, 2026 in Artificial Intelligence

Tencent has released Hy4 flash, its first Flash-tier model, listed at lower per-token prices than its flagship Hy4. Hy4 flash completed an automated task for 0.038 yuan, about one twenty-fourth of the cost for Hy4.
Why it matters
The release shows Tencent offering a cheaper, smaller model alongside its flagship, which may matter to developers weighing cost against capability. Hy4 flash scored 52.2 on the automated task test against 56.2 for Hy4, so the price gap comes with a measurable performance gap.
Hy4 flash has 224 billion total parameters, activating 12 billion per inference, and supports multimodal input with a 1 million token context window, compared with Hy4's 770 billion total and 49 billion activated parameters. The source does not say how the benchmark scores were measured or whether independent testers have reproduced them.
The players
Tencent
Tencent is a Chinese technology conglomerate known for its social media, gaming and cloud services, and it has developed the Hy4 model family.
The details
Tencent lists Hy4 flash at 0.5 yuan per million input tokens, 2 yuan per million output tokens, and 0.02 yuan per million cache-hit tokens. A limited-time 50% discount applies to Hy4, which narrows the current price difference between the two models.
Timeline
November 16, 2026: Article reporting the Hy4 flash release and benchmarks is published.
Roadmap
Tencent's Hy4 flash follows the tiered approach in which AI developers pair a flagship model with a cheaper, smaller variant to serve different cost and capability needs. The release places Tencent alongside rivals already offering multiple model tiers in the competitive large language model market.
Developers and businesses choosing an AI model can now compare Hy4 flash's lower per-token prices against Hy4's higher benchmark scores. The 50% limited-time discount on Hy4 means the cost gap may shift for anyone deciding which model to use.
The takeaway
Choosing between a flagship and a Flash-tier model depends on whether a task's quality needs justify the higher price. Comparing per-token costs against benchmark scores for your own workload is the practical next step.
Further reading
For more on how companies are building and pricing these systems, see Artificial Intelligence.



