tencent/hy4-preview

Model release OpenRouter 模型 Source language: Chinese

OpenRouter 于 2026-08-28 收录该模型(收录时间不等于模型正式发布日期);上下文:1,048,576 tokens;输入价:$0.7506/百万 tokens;输出价:$2.2509/百万 tokens;定位(来源描述):Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of…

Full text

Tencent: Hy4 preview

tencent/hy4-preview

Model weights

Compare

Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of 770B total. It is designed for coding agents, complex tool-use workflows, and productivity tasks that require planning, context continuity, and sustained multi-step execution.

Modalities

In / out price

$0.834 / $2.501per 1M

Context

1.0M

Released

Aug 28, 2026

Compare

Providers

Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick: Balanced (price + speed), Nitro (fastest), or Floor (cheapest).

Latency / throughput

$0.8340$2.50100$0.04201.50s78 tok/s

100.00%

$0.8340$2.50100$0.04201.32s66 tok/s

100.00%

$0.8340$2.50100$0.04203.84s28 tok/s

100.00%

$0.8340$2.50100$0.04201.59s36 tok/s

100.00%

$0.8340$2.50100$0.04201.75s48 tok/s

100.00%

$1.0425$3.12625$0.05250.84s67 tok/s

100.00%

$0.8340$2.50100$0.04203.54s26 tok/s

99.51%

ProviderInput / Output /M

Input $0.834 / Output $2.501

78 tok/s100.00% uptime

Input $0.834 / Output $2.501

66 tok/s100.00% uptime

Input $0.834 / Output $2.501

28 tok/s100.00% uptime

Input $0.834 / Output $2.501

36 tok/s100.00% uptime

Input $0.834 / Output $2.501

48 tok/s100.00% uptime

Input $1.0425 / Output $3.12625

67 tok/s100.00% uptime

Input $0.834 / Output $2.501

26 tok/s99.51% uptime

Pricing

The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.

Weighted average

Input

$0.0832/M tokens

Output

$2.508/M tokens

Hy4 preview — Price history

Price history

No effective price data for this selection

$0.0791$2.5095.3%68.8%

$0.0826$2.5094.9%16.0%

$0.0851$2.5094.6%7.9%

$0.1082$2.5091.6%2.4%

$0.1186$2.5090.3%4.3%

$0.1226$3.1392.9%0.3%

$0.2603$2.5072.4%0.3%

ProviderInput / Output /M

Token share

Effective in $0.0791 / Effective out $2.50

Token share

68.8%

Effective in $0.0826 / Effective out $2.50

Token share

16.0%

Effective in $0.0851 / Effective out $2.50

Token share

7.9%

Effective in $0.1082 / Effective out $2.50

Token share

2.4%

Effective in $0.1186 / Effective out $2.50

Token share

4.3%

Effective in $0.1226 / Effective out $3.13

Token share

0.3%

Effective in $0.2603 / Effective out $2.50

Token share

0.3%

Performance

Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).

Throughput

78tok/s

P50, best across providers

Latency

0.84s

P50, best provider

Reasoning effortLatency / throughput

AutoExacto benchmarks

GPQA DiamondTAU-BenchDeepInfra90.5%77.1%Venice89.9%76.7%auto-routing89.7%76.7%Cohere90.0%76.0%NovitaAI89.2%76.6%

+5 more providers

Uptime

Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.

Uptime (3d)The model was reachable. Request routed to a provider.

100.00%

Availability (3d)The model returned inference from any provider. Errors and empty responses count against it.

99.77%

Availability over the last 3 days

Last 72 hours

Availability 99.77%

3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability

99.59%

Without Routing

97.85%

When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and provider routing options.

Benchmarks

Scores on standardized evaluations and this model’s own outputs on shared benchmark prompts. Higher percentages are better — and each rank shows where this model lands among the models evaluated on that source.

Benchmark score summary for Tencent: Hy4 preview (Vals AI)

SourceBenchmarkScore

Vals AIHy4 Preview BioMysteryBench69.3%

Vals AIHy4 Preview Code Migration v1.145.4%

Vals AIHy4 Preview CyberBench v1.165.4%

Vals AIHy4 Preview EMB v1.168.9%

Vals AIHy4 Preview Finance Agent v2.161.6%

Vals AIHy4 Preview Vals Finance Index57.5%

Vals AIHy4 Preview HLAB-Vals-1.18.8%

Vals AIHy4 Preview IOI59.3%

Vals AIHy4 Preview LegalBench83.8%

Vals AIHy4 Preview Vals Legal Index71.7%

Vals AIHy4 Preview Legal Research Bench45.2%

Vals AIHy4 Preview MedCode43.2%

Vals AIHy4 Preview MedScribe83.6%

Vals AIHy4 Preview ProgramBench0.0%

Vals AIHy4 Preview ProofBench v1.175.0%

Vals AIHy4 Preview Public Benefits Bench v1.168.6%

Vals AIHy4 Preview SRE Bench2.3%

Vals AIHy4 Preview Tax Agent Bench32.6%

Vals AIHy4 Preview Terminal-Bench 4.08.1%

Vals AIHy4 Preview Terminal-Bench-Science1.4%

Vals AIHy4 Preview Vals Index53.2%

Vals AIHy4 Preview Vibe Code Bench v1.177.5%

Apps

Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.

1.

Claude Code

Claude Code is Anthropic's agentic coding tool that reads your entire codebase, plans and executes changes across files, runs tests, and iterates on failures, all from natural language prompts.

1.84Ttokens

2.

Hermes Agent

Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across sessions, and builds reusable skills from experience. It comes with 40+ built-in tools, including web search, browser automation, and vision, plus scheduled automations and subagents.

41.8Btokens

3.

Portkey AI

Control panel for AI apps

9.22Btokens

4.

One API

AI key management system

7Btokens

5.

Open Design

new

1.89Btokens

Top 5 Apps

1.85B

Rest of OpenRouter

930M

Activity

Token volume and request traffic to this model over time.

API

Drop-in code to call this model. OpenRouter's API is OpenAI-compatible — most SDKs work by just swapping the base URL. The only thing that changes between models is the model slug below.

Not currently available through EU in-region routing. In-region routing docs

Frequently asked questions

Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of 770B total. It is designed for coding agents, complex tool-use workflows, and productivity tasks that require planning, context continuity, and sustained multi-step execution.

Hy4 preview costs $0.834/M input tokens and $2.501/M output tokens, with separate rates for Cache Read at $0.042/M tokens.

Hy4 preview has a 1,048,576 token context window.

Yes. Hy4 preview accepts tools and tool_choice for function calling. It also supports structured outputs via a JSON schema in response_format.

Hy4 preview is served by 7 providers on OpenRouter: GMICloud, AtlasCloud, DeepInfra, NovitaAI, SiliconFlow, Tencent Cloud, and Venice. Requests are routed to the best available provider, with automatic failover to the others, and you can pin or exclude providers with provider routing.

Hy4 preview was released on August 28, 2026.

Sources

← Back to the daily brief