Inference
EN
Tencent

Hy4

Hy4 is a reasoning model from Tencent. 14 benchmarks count toward its index, in 4 categories.

  • New
  • Reasoning
  • Input: text · image
  • Released 2026-11
RankedEnough results for an index.
IndexAll categories, out of 100.72.0±7.8Median 51.0
CoverageIndex weight that has results.65%
SpeedOutput tokens per second.52/sMedian 67/s
First tokenSeconds to the first output token.3.4 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.417Median $0.60
Output / 1MUS dollars per 1M output tokens.$1.25Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.—Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
tencent/hy4
Max outputMaximum output tokens per request.
66K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.021
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
2 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Novita AIHost$0.417$1.251M
OpenRouterRouter$0.417$1.251.05M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.9 counted74.510 / 218
CodingCode writing and repair.3 counted68.424 / 209
ReasoningLogic problems and puzzles.None counted——
MultimodalTasks with images and text.1 counted57.4—
KnowledgeFacts and expert knowledge.1 counted83.7—
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.None counted——
MathMath problems.None counted——

Results

14 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
APEX-AgentsAgentic75.2%1 / 13
Humanity's Last ExamKnowledge59.1%7 / 77
AutomationBenchAgenticAbove fitFar above this model's index.56.2%4 / 32
DeepSWEAgentic74.9%8 / 59
Agent Arena task outcomeAgentic6.712 / 54
Vibe Code Bench v1.1Coding81.0% ±3.530 / 123
Terminal-Bench 4.0Agentic39.4% ±0.915 / 61
Agent Arena steerabilityAgentic8.814 / 54
Humanity's Last Exam with toolsAgentic59.1%10 / 36
FrontierCode 2.0Coding31.0%9 / 27
Agents' Last ExamAgentic41.0%8 / 21
Massive Multi-discipline Multimodal Understanding ProMultimodal78.6%22 / 53
OSWorld 2.1 (all 108 tasks)Agentic78.9%5 / 11
ProgramBench: Can Language Models Rebuild Programs From Scratch?Coding30.0%12 / 19
Terminal-Bench-Science 0.1Agentic19.8%11 / 11
Tax Agent BenchFinance34.4% ±3.317 / 82
Legal Research BenchLegal41.3% ±3.430 / 90
HLAB-Vals-1.1Legal12.5% ±2.717 / 35
ZeroBench_main with PythonMultimodal55.4%—
SWE-MarathonCoding43.8%—
PostTrainBench v1.2Coding35.2%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseVals AINo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index72.0 → 72.0
11 Nov 202612 Nov 2026
Output / 1M$1.25 → $1.25
3 Nov 202612 Nov 2026

Questions

How much does Hy4 cost?
$0.417 per 1M input tokens, $1.25 per 1M output tokens.
How good is Hy4?
Its index is 72.0 out of 100. That is place 19 of 283 ranked models.
What is Hy4 best at?
Agentic: 74.5 out of 100, place 10 of 218.
How fast is Hy4?
52 output tokens per second, 3.4 s to the first token.
What is the context window of Hy4?
1.05M tokens.
When was Hy4 released?
November 2026.

Badge

Hy4 on Inference360: place 19 of 283, index 72.0
HTML
<a href="https://inference360.ai/models/hy4"><img src="https://inference360.ai/badges/hy4.svg" alt="Hy4 on Inference360: place 19 of 283, index 72.0" height="20"></a>
Markdown
[![Hy4 on Inference360: place 19 of 283, index 72.0](https://inference360.ai/badges/hy4.svg)](https://inference360.ai/models/hy4)

Compare with

GPT-6 Sol72.8GPT-5.6 Sol73.0Muse Wonder73.1Claude Opus 573.8MiMo-V2.6-Pro70.2

Lower price

MiniMax M3.1 Flash69.3 · $0.60/1MMiMo-V2.6-Pro70.2 · $0.87/1M

More from Tencent

All models
Hy-Memory-30B-A3BUnrankedHy4 preview66.4Hy MT2 7BUnrankedAuKUnrankedUI-Mate-27BUnranked