Inference
EN
Tencent

Hy3 Preview

Hy3 Preview is a reasoning model from Tencent in the Hy3 family. 11 benchmarks count toward its index, in 5 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-04
RankedEnough results for an index.
IndexAll categories, out of 100.52.3±7.4Median 51.0
CoverageIndex weight that has results.75%
SpeedOutput tokens per second.69/sMedian 67/s
First tokenSeconds to the first output token.2.0 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.18Median $0.60
Output / 1MUS dollars per 1M output tokens.$0.60Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.1401±7 #155Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
tencent/hy3-preview
Max outputMaximum output tokens per request.
236K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
No
Cache read / 1MUS dollars per 1M cached tokens.
$0.06
WeightsLink to the published weights.
Hugging Face (opens in a new tab) ↗
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
1 provider
ProviderTypeMaker, cloud, host or router.InputOutputContext
OpenRouterRouter$0.18$0.60262K

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

Run locally

298.79B parameters
16-bit
628.5 GB
8-bit
334.3 GB
4-bit
191.2 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy needs more than 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.3 counted52.6115 / 218
CodingCode writing and repair.2 counted55.294 / 209
ReasoningLogic problems and puzzles.2 counted49.7118 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.3 counted53.9103 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted48.0—
MathMath problems.None counted——

Results

11 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Artificial Analysis GPQA DiamondKnowledge89.7%39 / 199
Graduate-Level Google-Proof Q&AKnowledge87.2%43 / 141
Artificial Analysis Humanity's Last ExamKnowledge33.5%68 / 222
GDPval-AA normalizedAgentic35.8%49 / 133
Critical Physics TasksReasoning4.9%84 / 214
Software Engineering Benchmark VerifiedCoding74.4%48 / 114
Artificial Analysis Omniscience AccuracyKnowledge31.5%75 / 164
Artificial Analysis SciCodeCoding48.6%55 / 114
Scientific Code BenchmarkCoding41.2%18 / 35
GPQA Diamond · BenchLMKnowledge87.2%48 / 84
Artificial Analysis Long Context ReasoningReasoning66.7%132 / 217
Terminal-Bench 2.0 · BenchLMAgentic54.4%25 / 41
Humanity's Last ExamKnowledge25.5%58 / 77
Instruction Following BenchmarkInstruction63.1%43 / 57
Gert Labs Composite Game BenchmarkAgentic36.9%43 / 52

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reusemodels.devMIT

History

5 releases
Index53.0 → 52.3
23 Oct 202612 Nov 2026
Output / 1M$0.60 → $0.60
23 Oct 202612 Nov 2026

Questions

How much does Hy3 Preview cost?
$0.18 per 1M input tokens, $0.60 per 1M output tokens.
How good is Hy3 Preview?
Its index is 52.3 out of 100. That is place 124 of 283 ranked models.
What is Hy3 Preview best at?
Knowledge: 53.9 out of 100, place 103 of 268.
How fast is Hy3 Preview?
69 output tokens per second, 2.0 s to the first token.
What is the context window of Hy3 Preview?
262K tokens.
Is Hy3 Preview open source?
Yes. Its weights are published.
When was Hy3 Preview released?
April 2026.

Badge

Hy3 Preview on Inference360: place 124 of 283, index 52.3
HTML
<a href="https://inference360.ai/models/hy3-preview"><img src="https://inference360.ai/badges/hy3-preview.svg" alt="Hy3 Preview on Inference360: place 124 of 283, index 52.3" height="20"></a>
Markdown
[![Hy3 Preview on Inference360: place 124 of 283, index 52.3](https://inference360.ai/badges/hy3-preview.svg)](https://inference360.ai/models/hy3-preview)

Compare with

Ripple-65B Preview52.3Kimi K2.552.2Ternary Bonsai 2 27B52.2dots3-note Preview63.4Hy4 preview66.4

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeNemotron 3 Ultra50.8 · Free

More from Tencent

All models
Hy-MT1.5-1.8B-1.25bitUnrankedHy-MT2-1.8BUnrankedHy-MT2-30B-A3BUnrankedHy3UnrankedHy ASR 3.0 previewUnranked