· via Vercel blog
Tencent's Hy4 Flash lands on Vercel AI Gateway with 50% introductory discount
Vercel has added Tencent's low-cost Hy4 Flash model to AI Gateway with 50% off all token types through November 29, 2026. The model retains the Hy4 family's full feature set at about one tenth of the per-token cost.

Tencent's budget model arrives on Vercel's gateway
Vercel has added Hy4 Flash, a lower-cost model from Tencent, to AI Gateway, its unified inference service. According to the Vercel blog, the model ships with an introductory 50% discount that applies to input, cached input, and output tokens and runs through November 29, 2026.
Full feature set at a lower price
Hy4 Flash is described as the cheaper sibling within Tencent's Hy4 family. According to Vercel, it retains the entire Hy4 capability set while costing about one tenth as much per token. That set includes reasoning that carries its thinking between conversation turns, tool calling, structured outputs, prompt caching, and image input, along with the same 1M-token context window as its pricier relatives.
Reasoning can also be switched off on a per-request basis. Vercel suggests this suits volume-heavy workloads that gain little from extended deliberation, naming coding, agent, and document-processing use cases as candidates.
How developers access it
On the gateway, the model is addressed as tencent/hy4-flash. For developers working inside coding agents such as Claude Code, Codex, or Cursor, Vercel recommends installing the latest Vercel CLI and running npx vercel ai-gateway setup, then selecting the model within the agent. It can also be tried directly in the company's model playground.
Vercel positions AI Gateway as a single API layer sitting in front of many model providers, adding usage and cost tracking, retries, failover, and performance tuning that the company says yields better uptime than calling providers directly. The service also offers custom reporting, Zero Data Retention support, spending budgets tied to API keys, and routing rules. Pricing on the gateway mirrors what providers charge, with no markup and no inference platform fee, including for Bring Your Own Key requests.
Why it matters
Per-token cost is often the deciding factor in whether an AI feature scales beyond a demo. A model that preserves tool calling, structured outputs, a long context window, and caching while cutting per-token prices by an order of magnitude changes the arithmetic for exactly the workloads Vercel highlights: agents that make many calls, coding assistants processing large files, and pipelines that chew through documents.
The arrival also fits Vercel's broader positioning of AI Gateway as the default route to any provider's models. When a cheaper Tencent option appears on an API a developer already uses, switching is a one-line model-name change rather than a procurement exercise, and because the gateway adds no markup, the introductory discount passes through to the customer in full.
One caveat: the details here come from a single announcement by Vercel itself, and the claim that Flash matches the full Hy4 feature set at a tenth of the price is the company's own characterization. Builders weighing a migration will want to benchmark the model against their current picks before committing, particularly since the 50% offer expires on November 29, 2026.
- #tencent
- #vercel
- #ai-gateway
- #llm
- #developer-tools