Tencent releases Hy4 with 1.0M context and first visual understanding, priced at $0.417 / $1.251 per 1M tokens

Tencent has released the official version of Hy4, a language model with a 1.0M-token context window that adds visual understanding for the first time. It is listed on OpenRouter as of November 4, 2026, at $0.417 per 1M input tokens and $1.251 per 1M output tokens, marked as 50% off.

What Tencent claims

According to the model description, the official release delivers "comprehensive and substantial improvements" over Hy4 Preview. Tencent says the gains are strongest in agentic capabilities and in performance across a wide range of long-horizon, real-world productivity tasks. The release also introduces visual understanding for the first time in the Hy4 line.

The page lists no benchmark scores for Hy4, so the improvement claims are Tencent's own and are not backed by published figures here.

Pricing and specs

Item	Value
Context window	1.0M tokens
Release date	Nov 4, 2026
Input price (discounted)	$0.417 per 1M tokens
Output price (discounted)	$1.251 per 1M tokens
Cache read (discounted)	$0.021 per 1M tokens
Pre-discount input / output	$0.834 / $2.501 per 1M tokens
Pre-discount cache read	$0.042 per 1M tokens

The 50% discount applies to the Tencent Cloud endpoint, the only provider listed. The page does not state how long the discount will last.

Live performance

As read on November 4, 2026, the Tencent Cloud endpoint showed:

• Latency: 1.38s (P50)
• Throughput: 140 tokens per second (P50)
• Uptime: 100.00%

OpenRouter's own availability figure for the model over the last 24 hours was 60.00% on the same date. These are live statistics and will change. OpenRouter says that when an upstream provider errors, it can reroute to another healthy provider if the request filters allow. Tencent Cloud is currently the only provider listed for Hy4.

Hy4 is not currently available through OpenRouter's EU in-region routing.

Context: Hy4 Preview and the Hy line

Tencent: Hy4 preview is described as a mixture-of-experts model with 49B active parameters out of 770B total. It is aimed at coding agents, complex tool-use workflows, and productivity tasks that need planning, context continuity, and sustained multi-step execution. It has a 1.0M context window and is listed at $0.834 input / $2.501 output per 1M tokens. The page does not give a parameter count for the official Hy4 release.

Other Tencent models listed on the page include Hy3 (295B total parameters, 21B active, 256K context in its description, $0.132 / $0.528 per 1M tokens) and Hy3 preview (262K context, $0.18 / $0.60).

What this means

The two concrete changes from Hy4 Preview are the added visual understanding and the move from a preview label to an official release. Both come from Tencent's own description. Teams that tested the preview for agentic coding or long-document workflows now have a model with image input at the same 1.0M context length. Until benchmark results or third-party evaluations appear, the size of the claimed agentic gains cannot be verified. With one provider listed and a 50% promotional price, buyers should check both the discount terms and endpoint availability before committing production workloads.