Tencent logo

Tencent Hy3

Tencent · Released Jul 6, 2026

Open SourcecodingreasoningagentsCompare this model

$0.2

Input price / MTok

#5 of 27

$0.8

Output price / MTok

#5 of 27

262K

Context window

#21 of 28

78

SWE-bench Verified

#6 of 12

57.9

SWE-bench Pro

#7 of 9

90.4

GPQA Diamond

#5 of 19

Summary

Tencent Hy3 is a open-source language model from Tencent, released on Jul 6, 2026.

At $0.2 input and $0.8 output per million tokens, it ranks #5 of 27 for input price, placing it mid-range on input cost.

Its context window of 262K tokens ranks #21 of 28 for context depth in this dataset.

It targets coding, reasoning, agents workloads.

295B total / 21B active MoE (192 experts, top-8 routing); no-think plus low/high chain-of-thought modes. Apache 2.0.

Benchmarks

BenchmarkScoreMax%RankSource
SWE-bench Verified7810078.0%#6/12Tencent Hunyuan (self-reported in offici...
SWE-bench Pro57.910057.9%#7/9Tencent Hunyuan (self-reported in offici...
GPQA Diamond90.410090.4%#5/19Tencent Hunyuan (self-reported in offici...

Similar Models

Compare all 6 models

Models with the most shared benchmark coverage and closest scores to Tencent Hy3.

Tencent Hy3
AnthropicClaude Sonnet 4.6
AnthropicClaude Opus 4.8
DeepSeekDeepSeek V4 Pro
DeepSeekDeepSeek V4 Flash
OpenAIGPT-5.5
SWE-bench Verified#6 of 12
Real-world GitHub issue resolution
SWE-bench Pro#7 of 9
Professional software engineering tasks
GPQA Diamond#5 of 19
Graduate-level science reasoning

Compare Tencent Hy3 side by side

Pre-filled with 5 similar models

Open comparison