Noometry

Tencent, open weights

Hy3

Hy3 by Tencent ranks 79th of 354 ranked models on the Noometry Index as of October 2026, with a score of 44.2. Its strongest category is coding, where it ranks 63rd. API pricing starts at $0.0825 per million input tokens and $0.33 per million output tokens, with a 262K-token context window.

Last verified

Noometry rank
#79 of 354
Index score
44.2
Evidence
Confirmed 19 results
Provider
Tencent
Released
July 6, 2026
Weights
Open weights
Reasoning
Yes
Context window
262K
Max output
128K
Input price
$0.0825 / M
Output price
$0.33 / M
Blended price
$0.14 / M
Output speed
Not measured
Value
#18 of 219
Knowledge cutoff
Unknown
Input
text
Hugging Face
tencent/Hy3

Category scores

Each category score combines every public result we have in that category.

  1. Coding 46.8
  2. Reasoning 26.1
  3. Math 40.1
  4. Knowledge 40.8
  5. Multilingual 53.5
  6. Instruction Following 75.1
  7. Long Context 44.1
  8. Writing & Preference 62.2
CategoryScoreRankResults
Coding46.8#632
Reasoning26.1#1362
Math40.1#931
Knowledge40.8#1141
Multilingual53.5#651
Instruction Following75.1#701
Long Context44.1#751
Writing & Preference62.2#813

Strengths and weaknesses

Categories where Hy3 places highest and lowest among the models ranked in each, with its score against that category's median.

Strongest categories

CategoryScorevs medianRank
Coding46.8+8.1#63 of 340, top 19%
Multilingual53.5+6.1#65 of 297, top 22%
Instruction Following75.1+3.8#70 of 305, top 23%

Weakest categories

CategoryScorevs medianRank
Reasoning26.1+2.5#136 of 350, top 39%
Knowledge40.8+3.5#114 of 314, top 37%
Math40.1+3.6#93 of 327, top 29%

Closest competitors

The models ranked just above and below Hy3. When scores are this close, price and speed are often the better way to choose.

ModelRankScoreBlended $/MSpeed
Gemini 2.5 Pro#7545.0$3.445Compare
GPT-5.4 mini#7645.0$1.6910Compare
Amazon Nova Experimental Chat 26 02 10#7744.5——Compare
DeepSeek-V3.2-Exp#7844.3$0.2916Compare
Inkling#8044.1$2.57—Compare
Claude Sonnet 4.5#8144.1$685Compare
Chatgpt 4o Latest 20250326#8243.8—21Compare
ERNIE 5.1#8343.8——Compare

Sponsored placements are available on pages like this one. Advertise on Noometry

Benchmark results

Every published result we track, with its source. Bold rows are the ones used for ranking; where several exist we prefer independent runs over self-reported numbers.

Coding

BenchmarkScorePositionSettingSourceDate
LMArena WebDev1508#46 of 113, top 41%LMArena2026-10-08
LMArena WebDev1355LMArena2026-10-08
LMArena Coding1464#68 of 294, top 24%LMArena2026-10-08
LMArena Coding1422LMArena2026-10-08

Reasoning

BenchmarkScorePositionSettingSourceDate
NYT Connections (extended)17.2%Lech Mazur benchmarks
NYT Connections (extended)41.2%#66 of 91, top 73%highLech Mazur benchmarks
LMArena Hard Prompts1447#69 of 297, top 24%LMArena2026-10-08
LMArena Hard Prompts1413LMArena2026-10-08

Math

BenchmarkScorePositionSettingSourceDate
LMArena Math1424LMArena2026-10-08
LMArena Math1475#33 of 285, top 12%LMArena2026-10-08

Knowledge

BenchmarkScorePositionSettingSourceDate
LMArena Expert1460#62 of 273, top 23%LMArena2026-10-08
LMArena Expert1423LMArena2026-10-08

Multilingual

BenchmarkScorePositionSettingSourceDate
LMArena Non-English1426#65 of 297, top 22%LMArena2026-10-08
LMArena Non-English1390LMArena2026-10-08
LMArena Chinese1433LMArena2026-10-08
LMArena Chinese1493#53 of 285, top 19%LMArena2026-10-08
LMArena French1424LMArena2026-10-08
LMArena French1461#48 of 223, top 22%LMArena2026-10-08
LMArena German1439#58 of 231, top 26%LMArena2026-10-08
LMArena Japanese1392#67 of 211, top 32%LMArena2026-10-08
LMArena Korean1395#61 of 213, top 29%LMArena2026-10-08
LMArena Russian1432#64 of 283, top 23%LMArena2026-10-08
LMArena Russian1398LMArena2026-10-08
LMArena Spanish1456#42 of 226, top 19%LMArena2026-10-08
LMArena Spanish1399LMArena2026-10-08

Instruction Following

BenchmarkScorePositionSettingSourceDate
LMArena Instruction Following1426#65 of 298, top 22%LMArena2026-10-08
LMArena Instruction Following1383LMArena2026-10-08

Long Context

BenchmarkScorePositionSettingSourceDate
LMArena Longer Query1402LMArena2026-10-08
LMArena Longer Query1442#62 of 291, top 22%LMArena2026-10-08

Writing & Preference

BenchmarkScorePositionSettingSourceDate
LMArena Text1401LMArena2026-10-08
LMArena Text1439#64 of 297, top 22%LMArena2026-10-08
LMArena Creative Writing1339LMArena2026-10-08
LMArena Creative Writing1402#75 of 295, top 26%LMArena2026-10-08
LMArena Multi-Turn1436#77 of 295, top 27%LMArena2026-10-08
LMArena Multi-Turn1411LMArena2026-10-08

API pricing by provider

RouteInput $/MOutput $/MCached input $/MChecked
deepinfra$0.13$0.53$0.0332026-10-10
openrouter$0.0825$0.33$0.02062026-10-10

All Tencent API prices → Estimate your cost →

Compare Hy3

Other Tencent models

Frequently asked questions

How good is Hy3?

Hy3 by Tencent ranks 79th of 354 ranked models on the Noometry Index as of October 2026, with a score of 44.2. Its strongest category is coding, where it ranks 63rd. API pricing starts at $0.0825 per million input tokens and $0.33 per million output tokens, with a 262K-token context window.

How much does Hy3 cost?

Hy3 costs $0.0825 per million input tokens and $0.33 per million output tokens on openrouter, with cached input at $0.0206.

What is Hy3's context window?

Hy3 accepts up to 262K tokens of input and can write up to 128K tokens in one response.

Is Hy3 open source?

Yes. Hy3's weights are downloadable from Hugging Face (tencent/Hy3); check the license for commercial terms.

What are Hy3's strengths and weaknesses?

Relative to other ranked models, Hy3 places best in coding, multilingual, instruction following and lowest in reasoning, knowledge, math.

What is Hy3 best at?

Its best category is coding, where it ranks 63rd on Noometry.