The LLM benchmark aggregator.

Hy3

Tencent · 299B · open weights

Strongest in Harm refusal (#3 of 300, on 1 of its 2 boards), weakest in Toxicity avoidance (#120 of 272). Above par in 10 of 12 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Opus 4.7 and ahead of Gemini 3.6 Flash and Qwen3.7 Max.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Safety60.3−1.2#4/3371/3

Harm refusal60.8−0.8#3/3001/2

Jailbreak resistance62.9−2#10/2721/1

Fairness64−6#26/3001/2

Secure code62.1−4.4#46/2741/1

Toxicity avoidance54.6−4.4#120/2721/1

Human preference62.5−5#51/3421/1

Agents53.6−14.5#79/2682/5

Knowledge work54.7−18.9#56/1782/2

Core abilities49.7−17.7#101/2041/3

General intelligence49.6−20.5#103/2041/3

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Hy3, left for the other.

Phi-4 Mini Instructwon 8 · lost 0

Magistral Smallwon 8 · lost 0

Jamba Mini 1.7won 8 · lost 0

Jamba Large 1.7won 8 · lost 0

Gemma 2B Itwon 8 · lost 0

Smollm2 1.7B Instructwon 8 · lost 0

Phi 3 Mini 128k Instructwon 8 · lost 0

Zephyr 7B Betawon 8 · lost 0

Gemma 7B Itwon 8 · lost 0

Phi 3 Mini 4k Instructwon 8 · lost 0

Nous Hermes 2 Mixtral 8x7b Dpowon 8 · lost 0

DeepSeek LLM 67B Chatwon 8 · lost 0

Qwen1.5 14B Chatwon 8 · lost 0

DBRX Instructwon 8 · lost 0

GPT-3.5 Turbowon 8 · lost 0

Llama 3 8B Instructwon 8 · lost 0

Jamba 1.5 Miniwon 8 · lost 0

Nova Litewon 8 · lost 0

Gemma 2 9B Itwon 8 · lost 0

Mistral Small 24B Instructwon 8 · lost 0

Command R Pluswon 8 · lost 0

Jamba 1.5 Largewon 8 · lost 0

Gemini 1.5 Flashwon 8 · lost 0

QwQ 32Bwon 8 · lost 0

Gemini 1.5 Prowon 8 · lost 0

Gemini 2.0 Flash Litewon 8 · lost 0

Gemini 2.5 Flash Litewon 8 · lost 0

Qwen3 Vl 235B A22B Instructwon 8 · lost 0

Mimo V2.6 Flashwon 8 · lost 0

Gemma 4 12Bwon 9 · lost 0

Llama 3.2 3B Instructwon 9 · lost 0

Nova Prowon 9 · lost 0

Llama 3.1 8B Instructwon 10 · lost 0

GPT-4won 10 · lost 0

Magistral Mediumwon 10 · lost 0

Gemini 2.0 Flashwon 10 · lost 0

O3 Miniwon 10 · lost 0

Gemma 3 4B Itwon 11 · lost 0

Command Awon 11 · lost 0

Mistral Smallwon 11 · lost 0

GPT-4.1 Nanowon 12 · lost 0

GPT-4.1 Miniwon 12 · lost 0

Grok 4.3won 12 · lost 0

GPT-4o Miniwon 11 · lost 1

GPT OSS 20Bwon 11 · lost 1

Llama 4 Scout Instructwon 11 · lost 1

GPT OSS 120Bwon 11 · lost 1

Gemma 3 27B Itwon 11 · lost 1

Qwen3.5 122B A10Bwon 11 · lost 1

GLM-4.6won 11 · lost 1

Nemotron 3 Ultra 550B A55Bwon 9 · lost 1

Claude 3 Haikuwon 9 · lost 1

Llama 3 70B Instructwon 9 · lost 1

Gemma 3 12B Itwon 9 · lost 1

Gemma 4 26B A4Bwon 9 · lost 1

GPT-5.1won 9 · lost 1

Llama 3.2 1B Instructwon 8 · lost 1

Granite 3.1 8B Instructwon 8 · lost 1

Nova Microwon 8 · lost 1

Phi 4won 8 · lost 1

Mistral Largewon 8 · lost 1

Qwen3 30B A3B Instructwon 8 · lost 1

Grok 4.1 Fastwon 8 · lost 1

Gemma 4 E2Bwon 7 · lost 1

Qwen3 8Bwon 7 · lost 1

Ling 3.0 Flashwon 7 · lost 1

Mistral 7B Instruct V0.2won 7 · lost 1

Granite 3.0 2B Instructwon 7 · lost 1

Phi 3 Small 8k Instructwon 7 · lost 1

Starling Lm 7B Betawon 7 · lost 1

Granite 3.1 2B Instructwon 7 · lost 1

Granite 3.0 8B Instructwon 7 · lost 1

Phi 3 Medium 4k Instructwon 7 · lost 1

Gemma 2 2B Itwon 7 · lost 1

Llama 3.1 Tulu 3 8Bwon 7 · lost 1

C4ai Aya Expanse 8Bwon 7 · lost 1

Qwen2 72B Instructwon 7 · lost 1

C4ai Aya Expanse 32Bwon 7 · lost 1

Claude 3 Sonnetwon 7 · lost 1

Qwen2.5 72B Instructwon 7 · lost 1

Qwen Maxwon 7 · lost 1

GPT-4 Turbowon 7 · lost 1

GPT-4owon 7 · lost 1

O1 Miniwon 7 · lost 1

Qwen3.5 27Bwon 7 · lost 1

DeepSeek V3won 10 · lost 2

Claude Haiku 4.5won 10 · lost 2

Gemini 2.5 Prowon 10 · lost 2

Gemma 4 31Bwon 10 · lost 2

Kimi K2.7 Codewon 8 · lost 2

Grok Build 0.1won 8 · lost 2

NVIDIA Nemotron 3 Super 120B A12Bwon 8 · lost 2

Qwen3.5 35B A3Bwon 8 · lost 2

GLM-4.5won 8 · lost 2

Grok 3won 8 · lost 2

Qwen3 Next 80B A3B Thinkingwon 7 · lost 2

Claude 3.7 Sonnetwon 7 · lost 2

Gemma 4 E4Bwon 6 · lost 2

Gemma 2 27B Itwon 6 · lost 2

O1won 6 · lost 2

Llama 4 Maverick Instructwon 9 · lost 3

GPT-5 Miniwon 9 · lost 3

Claude Sonnet 4won 9 · lost 3

DeepSeek R1won 9 · lost 3

DeepSeek V3.2won 9 · lost 3

DeepSeek V4 Flashwon 9 · lost 3

MiniMax M3won 9 · lost 3

Inklingwon 9 · lost 3

GPT-5.6 Lunawon 9 · lost 3

Kimi K2.6won 9 · lost 3

O4 Miniwon 8 · lost 3

Gemini 2.5 Flashwon 8 · lost 3

Grok 4won 8 · lost 3

GPT-4.1won 8 · lost 3

O3won 8 · lost 3

Ling 3.0 Flash VLwon 7 · lost 3

Claude 3.5 Haikuwon 7 · lost 3

Claude Opus 4won 7 · lost 3

GPT-6 Lunawon 8 · lost 4

Kimi K2.5won 8 · lost 4

GPT-6 Solwon 8 · lost 4

GPT-5 Nanowon 7 · lost 4

Kimi K2won 7 · lost 4

Qwen3 235B A22B Instructwon 7 · lost 4

GPT-5.2won 7 · lost 4

Kimi K2 Thinkingwon 5 · lost 3

Claude 3 Opuswon 5 · lost 3

MiniMax M2won 6 · lost 4

MiniMax M2.7won 6 · lost 4

GLM-5.1won 6 · lost 4

Claude Sonnet 4.6won 6 · lost 4

GPT-5won 7 · lost 5

Qwen3.5 397B A17Bwon 7 · lost 5

Mimo V2.5 Prowon 7 · lost 5

Qwen3.7 Maxwon 7 · lost 5

Gemini 3.6 Flashwon 7 · lost 5

Claude 3.5 Sonnetwon 4 · lost 4

Claude Opus 4.6won 4 · lost 4

Grok 4.6won 6 · lost 6

DeepSeek V4 Prowon 6 · lost 6

GLM-5.2won 6 · lost 6

Gemini 3.5 Flashwon 6 · lost 6

Grok 4.7won 5 · lost 7

Claude Sonnet 4.5won 5 · lost 7

Grok 4.5won 5 · lost 7

GPT-5.6 Terrawon 5 · lost 7

GLM-5.3 Flashwon 5 · lost 7

Qwen3.8 Maxwon 5 · lost 7

Mimo V2.6 Prowon 5 · lost 7

GPT-5.5won 5 · lost 7

GPT-5.6 Solwon 5 · lost 7

Muse Spark 1.1won 5 · lost 7

Gemini 3.8 Flashwon 5 · lost 7

Gemini 3 Prowon 3 · lost 5

GPT-5.4won 4 · lost 8

Claude Opus 4.8won 4 · lost 8

Claude Sonnet 5won 3 · lost 9

Claude Opus 5won 3 · lost 9

Claude Opus 4.7won 3 · lost 9

Claude Opus 5.5won 2 · lost 10

§ 3 · Sources

Where the numbers come from

5 publications, 10 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26

LMArena Text 1457

Artificial Analysis ↗1 measureread 2026-09-26

Artificial Analysis Intelligence Index 25

GDPval-AA ↗1 measureread 2026-09-26

GDPval-AA 1045

AA-Briefcase ↗1 measureread 2026-09-26

AA-Briefcase 944

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26

Enkrypt · Jailbreak risk 1.9%Enkrypt · Harmful content risk 7.2%Enkrypt · CBRN risk 0.5%Enkrypt · Toxicity risk 3.1%Enkrypt · Bias risk 56.1%Enkrypt · Insecure code risk 8.9%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=hy3)](https://publicai.io/model-index/m/hy3)