Model · Tencent

Tencent Hy4

Tencent Hy4 is served by Tencent through OpenRouter. It provides 1.05M tokens of context and supports text, vision inputs.

Snapshot 2026-11-06 tencent/hy4 Edited by the Hunter Alpha Hub desk

Context window

1.05M tokens

Input / 1M tokens

$0.417

Output / 1M tokens

$1.25

Key facts

Provider	Tencent
OpenRouter ID	`tencent/hy4`
Modalities	Text, Vision
Best for	Agents, Long Context, Multimodal

Source openrouter.ai/api/v1/models Verified 2026-11-06 Drift-checked daily 06:00 CST

Our read on Tencent Hy4

Tencent Hy4's headline is the 1,048,576-token window, but the number most likely to decide an agent's bill is the cache read: $0.021 per million against $0.417 for fresh input. That roughly 20x gap only matters if your harness really re-sends the same instructions, tool definitions and project context on every step. If it rebuilds the prompt each time, you pay the higher number and the attractive cache line is irrelevant. Measure the hit rate before treating Hy4 as a cheap long-context model.

This is not the budget row on the page. At $0.417 / $1.2505 per million it costs more than GLM 5.3 Flash and DeepSeek V4.1 Flash in this snapshot, and its 65,536-token output ceiling is half what several frontier models allow. What the premium buys is a first-party Tencent endpoint with image input, tools, structured outputs and explicit reasoning-effort controls. The catalogue's official-release note claims a substantial agentic improvement over Hy4 Preview; that claim is the vendor's, not a test result from this site.

Choose it when

Long-running tool loops with a stable cached prefix, screenshot or diagram input, and a workload where one first-party Tencent endpoint is easier to account for than a multi-provider route.

Skip it when

The job is high-volume text at the lowest possible token price, needs audio or video input, or regularly produces more than 65,536 output tokens.

Closest call. Z.ai GLM 5.3 Flash : The catalogue's GLM reading is dramatically cheaper per token and adds video input, but it has moved repeatedly. Use Hy4 only if your own tests show a gain that survives that price gap and the lower output ceiling.

Price, context, output limit, cache read, modalities, endpoint count and supported parameters were read from the OpenRouter catalogue on 2026-11-06. The catalogue also lists an Artificial Analysis intelligence index of 34.6 but no coding or agentic subindex. No private benchmark was run for this note.

Cost estimates

These examples use the 2026-11-06 input and output prices. Cache discounts, provider surcharges and tool fees are excluded.

Workload	Input	Output	Estimated cost
Quick test	20,000	5,000	$0.015
Product chat month	500,000	125,000	$0.365
Document batch	5,000,000	1,000,000	$3.34

Strengths

✓ A 1,048,576-token context with text and image input at $0.417 / $1.2505 per million
✓ Cached reads at $0.021 per million — about a twentieth of the fresh-input price, and the number that matters for repeated agent prefixes
✓ The endpoint declares tools, structured outputs and reasoning effort settings, so it can be wired into ordinary agent loops without a custom protocol
✓ The official release positions Hy4 as a substantial agentic improvement over Hy4 Preview

Limitations

! Text output only, with image input but no audio, video or generic file input
! The 65,536-token output ceiling is half the 128K allowance on several frontier rows here
! One Tencent endpoint and one fp4 serving path, so there is no provider diversity to fall back to inside the model id
! The catalogue carries a single Artificial Analysis intelligence index of 34.6 and no coding or agentic subindex; that is not enough to rank it against the frontier on a real workload

OpenRouter API quickstart

Run these examples from your server or edge function. Keep the OpenRouter API key private and do not expose it in browser code.

cURL

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tencent/hy4",
    "messages": [{ "role": "user", "content": "Summarize this document in five bullet points." }],
    "max_tokens": 500
  }'

JavaScript / TypeScript

const response = await fetch("https://openrouter.ai/api/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.OPENROUTER_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "tencent/hy4",
    messages: [{ role: "user", content: "Summarize this document in five bullet points." }],
    max_tokens: 500,
  }),
});

const data = await response.json();
console.log(data.choices[0].message.content);

Compare models View on OpenRouter ↗

Tencent Hy4 FAQ

What is Tencent Hy4?

Tencent Hy4 is a Tencent model available through OpenRouter. It supports text, vision inputs and is suited for agents, long context, multimodal workloads.

How much does Tencent Hy4 cost on OpenRouter?

The 2026-11-06 snapshot lists $0.417 per 1M input tokens and $1.25 per 1M output tokens. Verify current pricing on OpenRouter before production.

How large is Tencent Hy4's context window?

Tencent Hy4 offers 1.05M tokens of context in the current snapshot. Provider-specific limits and surcharges may apply.

Head-to-head comparisons

Tencent Hy4 vs Z.ai GLM 5.3 Flash

Start with GLM 5.3 Flash when its current catalogue price holds and video input matters. Pay for Hy4 only after your own agent or document tests show a reason: its 65K output ceiling is lower, but its cache-read price is $0.021 per million and its single Tencent endpoint makes the serving path easier to account for.

Pricing and context data are based on a 2026-11-06 snapshot. Provider limits can change; always verify on OpenRouter before production.