Model

Hunyuan-T1

CLOSED

Tencent · Hunyuan family · released Mar 21, 2025

Mamba-powered deep-reasoning flagship built on TurboS, API-only via Tencent Cloud/Yuanbao; scores 87.2 MMLU-Pro and 96.2 MATH-500, close to DeepSeek R1 and trailing only o1.

ReasoningCodingVisionFunction callingTool useAgentic

1631.0

Elo · rank #96

Parameters

Undisclosed

Active params

Undisclosed

Context

256K tokens

Architecture

Hybrid Transformer-Mamba MoE deep-reasoning model built on TurboS base (params undisclosed)

License

Proprietary

Languages

—

API price (in/out)

$0.15 / $0.59

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

AIMEMath78.2%#83

best: GPT-5.2 · 100.0%

Arena-HardHuman preference91.9%#6

best: Qwen3-235B-A22B (Non-Thinking) · 96.1%

C-EvalKnowledge91.8%#10

best: Qwen3.6-Plus · 93.3%

CMMLUKnowledge90.0%#4

best: Doubao-1.5-Pro · 90.9%

DROPReasoning93.1#1

best: this model · 93.1

GPQA DiamondReasoning69.3%#126

best: GPT-5.6 · 94.6%

LiveCodeBenchCoding64.9%#85

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MATH-500Math96.2%#26

best: GPT-5 · 99.4%

MMLU-ProKnowledge87.2%#12

best: Claude Fable 5 · 91.5%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$0.15

Output / M tok

$0.59

Hunyuan family

Elo progression across releases

Hunyuan (1st-gen)Sep 20231000.0 Hunyuan-Large (A52B)Nov 20241072.5 Hunyuan-TurboSFeb 20251306.0 Hunyuan-T1Mar 20251631.0 Hunyuan-A13BJun 20251517.1 Hy3Jul 20262485.4

Resources

DOCStencent.github.io/llm.hunyuan.T1/README_EN.html APIAvailable via Tencent API

API price $0.15/$0.59 · each benchmark row carries its own source badge (see methodology)