Skip to main content

Model profile

Hy3 Preview

TencentSupersededReleased Apr 23, 2026
Data verified
Superseded:Tencent has released newer models in this line —Hy3
Overall Score
Unranked
Arena Elo
1412
Eligible category ranks
3of 8
Price (1M tokens)
$0 in / $0 out
API pricing
Speed
Not listed
Context
256K

Evidence coverage

22 of 323 tracked benchmarks are published. 6 are verified and 16 provisional. 5 of 8 categories are measured.

Updated July 23, 2026Methodology
Published / tracked
22 / 323
Verified
6
Provisional
16
Categories with evidence
5 / 8

Evidence by category

  • Agentic5 benchmarks
    Mixed evidence
  • Coding5 benchmarks
    Mixed evidence
  • Reasoning2 benchmarks
    Reported
  • Knowledge9 benchmarks
    Mixed evidence
  • Math0 benchmarks
    Not measured
  • Multilingual0 benchmarks
    Not measured
  • Multimodal0 benchmarks
    Not measured
  • Inst. Following1 benchmark
    Reported
Open WeightSelf-hostReasoning
Confidence:
Medium
preview

BenchLM is tracking Hy3 Preview, but this profile is currently excluded from the public leaderboard because it still lacks enough non-generated benchmark coverage to rank safely. Only non-generated public benchmark rows appear below.

Hy3 Preview is a open weight model with a 256K token context window. It uses explicit chain-of-thought reasoning, which typically improves performance on math and complex reasoning tasks at the cost of higher latency and token usage.

Hy3 Preview sits inside the Hy3 family alongside Hy3. This profile currently has 22 of 323 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Its strongest category is Instruction Following (#22), while its weakest is Coding (#65). This performance profile makes it a well-rounded choice across a range of tasks.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 77.4483.93

  1. Claude Mythos 5
    Anthropic
    #183.93
    Claude Mythos 5 is #1 with a score of 83.93.
    Compare
  2. Claude Fable 5
    Anthropic
    #283.68
    Claude Fable 5 is #2 with a score of 83.68.
    Compare
  3. GPT-5.6 Sol
    OpenAI
    #381.96
    GPT-5.6 Sol is #3 with a score of 81.96.
    Compare
  4. Kimi K3
    Moonshot AI
    #480.96
    Kimi K3 is #4 with a score of 80.96.
    Compare
  5. Claude Opus 4.8
    Anthropic
    #578.34
    Claude Opus 4.8 is #5 with a score of 78.34.
    Compare
  6. Muse Spark 1.1
    Meta
    #677.44
    Muse Spark 1.1 is #6 with a score of 77.44.
    Compare
  7. Hy3 PreviewCurrent model
    Tencent
    UnrankedNot measured
    Hy3 Preview is Unranked with a score of Not measured.

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

  1. Inst. Following30%
    Eligible cohort rank #22 of 31Category score 59.0
  2. Agentic60%
    Eligible cohort rank #48 of 119Category score 49.9
  3. Coding47%
    Eligible cohort rank #65 of 122Category score 49.6

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #48 of 119Percentile 60thWeight 22%5 benchmarksMixed sources49.9
CodingRank #65 of 122Percentile 47thWeight 20%5 benchmarksMixed sources49.6
ReasoningWeight 17%2 benchmarksReportedScore pending
KnowledgeRank Not rankedWeight 12%9 benchmarksMixed sources39.9
MathWeight 5%0 benchmarksNot measuredNot measured
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalWeight 12%0 benchmarksNot measuredNot measured
Inst. FollowingRank #22 of 31Percentile 30thWeight 5%1 benchmarkReported59.0

Chatbot Arena performance

Scroll horizontally to inspect confidence intervals and vote counts.

Chatbot Arena Elo, confidence interval, and vote count by evaluation view
ViewEloConfidence intervalVotes
Text Overall1412±7.56,637
Coding1459±13.71,965
Math1428±27.6405
Instruction Following1398±12.92,197
Creative Writing1355±19.41,007
Multi-turn1415±17.91,169
Hard Prompts1437±9.34,383
Hard Prompts (English)1437±12.82,211
Longer Query1425±11.62,875

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic5 benchmarks
Terminal-Bench 2.0Provider exact
54.4%Weighted 38%
Source: Tencent Hy3-preview model cardProvenance: Tencent/Hugging Face reports Terminal-Bench 2.0 at 54.4.
Gert LabsBenchmark exact

Gert Labs Composite Game Benchmark

36.91%Display only
Source: Gert Labs rankingsProvenance: Gert Labs reports this composite leaderboard score in the public rankings API. BenchLM scales the source gscore from 0-1 to 0-100 and stores it as a display-only agentic benchmark.
AA Agentic IndexReported

Artificial Analysis Agentic Index

30.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
GDPval-AASecondary exact

GDPval-AA normalized

35.7%Display only
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports normalized GDPval at 36.84%.
GDPval-AAReported
1214Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Coding5 benchmarks
SWE-bench VerifiedProvider exact

Software Engineering Benchmark Verified

74.4%Weighted 16%
Source: Tencent Hy3-preview model cardProvenance: Tencent/Hugging Face reports SWE-bench Verified resolved at 74.4.
SciCodeSecondary exact

Scientific Code Benchmark

41.2%Weighted 16%
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports SciCode at 41.2%.
Terminal-Bench 2.0Provider exact
54.4%Display only
Source: Tencent Hy3-preview model cardProvenance: Tencent/Hugging Face reports Terminal-Bench 2.0 at 54.4. BenchLM stores the value in both coding and agentic views because the benchmark is used as an agentic coding signal.
AA-SciCodeReported

Artificial Analysis SciCode

47.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA Coding IndexReported

Artificial Analysis Coding Index

58.8%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Reasoning2 benchmarks
AA-LCRSecondary exact

Artificial Analysis Long Context Reasoning

66.7%Display only
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports AA-LCR at 54.7%.
CritPtSecondary exact

Critical Physics Tasks

4.9%Display only
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports AA-CRIT-PT at 4.6%.
Knowledge9 benchmarks
HLESecondary exact

Humanity's Last Exam

25.5%Weighted 45%
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports Humanity's Last Exam at 25.5%.
GPQAProvider exact

Graduate-Level Google-Proof Q&A

87.2%Weighted 7%
Source: Tencent Hy3-preview model cardProvenance: Tencent/Hugging Face reports GPQA Diamond at 87.2. BenchLM stores that exact value on the weighted GPQA lane and the display GPQA-Diamond lane.
Artificial Analysis Intelligence IndexSecondary exact
41.2%Display only
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports Hy3-preview (Reasoning) at 41.85 on the Intelligence Index.
GPQA-DProvider exact

GPQA Diamond

87.2%Display only
Source: Tencent Hy3-preview model cardProvenance: Tencent/Hugging Face reports GPQA Diamond at 87.2.
AA-Omniscience AccuracySecondary exact

Artificial Analysis Omniscience Accuracy

31.5%Display only
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports AA-Omniscience Accuracy at 28.0%.
AA-Omniscience Hallucination RateSecondary exact

Artificial Analysis Omniscience Hallucination Rate

73.0%Display only
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports AA-Omniscience hallucination rate at 86.9%. Lower is better for this row.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

89.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

31.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

-18.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
IFBenchSecondary exact

Instruction Following Benchmark

63.1%Weighted 65%
Source: Artificial Analysis: Hy3-previewProvenance: Artificial Analysis reports IFBench at 63.1%.

Hy3 Family

preview · preview

Canonical Entry

Hy3

Frequently Asked Questions

How does Hy3 Preview perform overall in AI benchmarks?

Hy3 Preview has 22 published benchmark scores on BenchLM, but it does not yet have enough non-generated coverage to receive a global overall rank.

Is Hy3 Preview good for knowledge and understanding?

Hy3 Preview has visible benchmark coverage in knowledge and understanding, but BenchLM does not currently assign it a global category rank there.

Is Hy3 Preview good for coding and programming?

Hy3 Preview ranks #65 out of 122 models in coding and programming benchmarks with an average score of 49.6. There are stronger options in this category.

Is Hy3 Preview good for reasoning and logic?

Hy3 Preview has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is Hy3 Preview good for agentic tool use and computer tasks?

Hy3 Preview ranks #48 out of 119 models in agentic tool use and computer tasks benchmarks with an average score of 49.9. There are stronger options in this category.

Is Hy3 Preview good for instruction following?

Hy3 Preview ranks #22 out of 31 models in instruction following benchmarks with an average score of 59. There are stronger options in this category.

Is Hy3 Preview open source?

Yes, Hy3 Preview is an open weight model created by Tencent, meaning it can be downloaded and run locally or fine-tuned for specific use cases.

Which sibling models are related to Hy3 Preview?

Hy3 Preview belongs to the Hy3 family. Related variants on BenchLM include Hy3.

Does Hy3 Preview have full benchmark coverage on BenchLM?

Not yet. Hy3 Preview currently has 22 published benchmark scores out of the 323 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of Hy3 Preview?

Hy3 Preview has a published context window of 256K, which determines how much text it can process in a single interaction.

Last updated: July 23, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.