Tencent model product
Hy4
Hy4 is a Tencent preview Mixture-of-Experts (MoE) language model with 770B total parameters, 49B activated per token, and a 1,048,576-token context window.
Updated Sep 8, 2026. Default version: Hy4 preview
LLMBoard Score83.7Hy4 preview
Coverage20%30 benchmark families
Context windowN/ATokens
Official input priceN/AOfficial price unavailable
Hy4 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Hy4 Benchmark Results
Benchmark scores for Hy4 preview.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkBankerToolBench | Score78.60% | Rank01 | Participants2 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkCritPT | Score16.90% | Rank01 | Participants6 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkDRACO | Score77.20% | Rank01 | Participants2 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkHorizonMath | Score8.80% | Rank01 | Participants4 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkSuperChem | Score66.40% | Rank01 | Participants4 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkWideSearch | Score83.90% | Rank01 | Participants11 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkArXivMath | Score66.60% | Rank02 | Participants3 | Percentile50.00% | EvidenceC | Evaluated |
| BenchmarkBioMysteryBench | Score71.30% | Rank02 | Participants5 | Percentile75.00% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam (no tools, text-only) | Score43.40% | Rank02 | Participants3 | Percentile50.00% | EvidenceC | Evaluated |
| BenchmarkJob Bench | Score61.70% | Rank02 | Participants8 | Percentile85.71% | EvidenceC | Evaluated |
| BenchmarkNL2Repo | Score58.90% | Rank02 | Participants22 | Percentile95.24% | EvidenceC | Evaluated |
| BenchmarkSkillsBench | Score62.90% | Rank02 | Participants9 | Percentile87.50% | EvidenceC | Evaluated |
| BenchmarkWorkspace Bench | Score60.20% | Rank02 | Participants4 | Percentile66.67% | EvidenceC | Evaluated |
| BenchmarkAPEX-Agents | Score37.10% | Rank03 | Participants9 | Percentile75.00% | EvidenceC | Evaluated |
| BenchmarkMathArena Apex | Score74.20% | Rank03 | Participants9 | Percentile75.00% | EvidenceC | Evaluated |
| BenchmarkOneMillion Bench | Score65.40% | Rank03 | Participants4 | Percentile33.33% | EvidenceC | Evaluated |
| BenchmarkSWE-bench Multilingual | Score82.90% | Rank03 | Participants43 | Percentile95.24% | EvidenceC | Evaluated |
| BenchmarkSWE-Marathon | Score31.90% | Rank03 | Participants5 | Percentile50.00% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam (with tools, text-only) | Score55.40% | Rank04 | Participants5 | Percentile25.00% | EvidenceC | Evaluated |
| BenchmarkMCP Atlas | Score83.70% | Rank04 | Participants34 | Percentile90.91% | EvidenceC | Evaluated |
| BenchmarkOfficeQA Pro | Score66.20% | Rank04 | Participants9 | Percentile62.50% | EvidenceC | Evaluated |
| BenchmarkPostTrainBench | Score35.60% | Rank04 | Participants7 | Percentile50.00% | EvidenceC | Evaluated |
| BenchmarkToolathlon | Score74.10% | Rank04 | Participants41 | Percentile92.50% | EvidenceC | Evaluated |
| BenchmarkAutomationBench | Score32.10% | Rank05 | Participants18 | Percentile76.47% | EvidenceC | Evaluated |
| BenchmarkDeepSWE | Score64.30% | Rank05 | Participants13 | Percentile66.67% | EvidenceC | Evaluated |
| BenchmarkSWE-Bench Pro | Score65.70% | Rank05 | Participants55 | Percentile92.59% | EvidenceC | Evaluated |
| BenchmarkLM Arena Agent Task Outcome Explicit | Score13.56% | Rank05 | Participants49 | Percentile91.67% | EvidenceA | Evaluated |
| BenchmarkGDPval-AA | Score1,678.00 points | Rank06 | Participants9 | Percentile37.50% | EvidenceC | Evaluated |
| BenchmarkProgram Bench | Score17.50% | Rank07 | Participants7 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkCyberGym | Score78.40% | Rank08 | Participants15 | Percentile50.00% | EvidenceC | Evaluated |
Hy4 Arena Results
Preference and agent-evaluation results for the default version.
Sort by
Arena | Category | Rank | Rating / score | Votes | Observations | Result date |
|---|
| Arenaagent task outcome explicit | Categoryoverall | Rank05 | Rating / score0.14 | VotesN/A | Observations6.5K | Result date |
| Arenawebdev | Categorywebdev-react | Rank07 | Rating / score1,640.93 | Votes1,134 | ObservationsN/A | Result date |
| Arenaagent bash recovery steps | Categoryoverall | Rank08 | Rating / score0.10 | VotesN/A | Observations36.2K | Result date |
| Arenaagent | Categoryoverall | Rank09 | Rating / score0.07 | VotesN/A | Observations1.2M | Result date |
| Arenawebdev | Categoryoverall | Rank10 | Rating / score1,621.26 | Votes1,661 | ObservationsN/A | Result date |
| Arenawebdev | Categorywebdev | Rank10 | Rating / score1,621.26 | Votes1,661 | ObservationsN/A | Result date |
| Arenaagent praise complaint | Categoryoverall | Rank11 | Rating / score0.10 | VotesN/A | Observations2.3K | Result date |
| Arenawebdev | Categorywebdev-html | Rank12 | Rating / score1,603.03 | Votes179 | ObservationsN/A | Result date |
| Arenaagent steerability | Categoryoverall | Rank24 | Rating / score0.00 | VotesN/A | Observations5.9K | Result date |
| Arenaagent tool hallucination | Categoryoverall | Rank28 | Rating / score0.01 | VotesN/A | Observations1.2M | Result date |
Hy4 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- From $0.67 input, $2 output per 1M via Vancine
- Tracked offerings
- 9
Sort by
Provider | Provider model ID | Region | Input / 1M | Output / 1M | Context | Updated |
|---|
| ProviderVancine | Provider model IDhy4-preview | Regionglobal | Input / 1M$0.67 | Output / 1M$2 | Context1M | Updated |
| ProviderNanoGPT | Provider model IDtencent/hy4-preview | Regionglobal | Input / 1M$0.834 | Output / 1M$2.5 | Context1M | Updated |
| ProviderTencent TokenHub | Provider model IDhy4-preview | Regionglobal | Input / 1M$0.834 | Output / 1M$2.5 | Context1M | Updated |
| ProviderOpenRouter | Provider model IDtencent/hy4-preview | Regionglobal | Input / 1M$0.834 | Output / 1M$2.5 | Context1M | Updated |
| ProviderOpenCode Go | Provider model IDhy4-preview | Regionglobal | Input / 1M$0.834 | Output / 1M$2.5 | Context1M | Updated |
| ProviderVercel AI Gateway | Provider model IDtencent/hy4-preview | Regionglobal | Input / 1M$0.834 | Output / 1M$2.5 | Context1M | Updated |
| ProviderTencent Token Plan | Provider model IDhy4-preview | Regionglobal | Input / 1M$0.834 | Output / 1M$2.5 | Context1M | Updated |
| ProviderKilo Gateway | Provider model IDtencent/hy4-preview | Regionglobal | Input / 1M$0.834 | Output / 1M$2.5 | Context1M | Updated |
| ProviderCrossModel | Provider model IDtencent/hy4-preview | Regionglobal | Input / 1M$0.96 | Output / 1M$2.88 | Context1M | Updated |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Hy4 Runtime Performance
Provider-specific output speed and catalog latency for Hy4 preview. Runtime does not affect the capability score.
No runtime data
No provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Hy4 Specifications
Technical details for the model's default version.
- Version
- Hy4 preview
- Released
- Aug 28, 2026
- Knowledge cutoff
- Unknown
- Parameters
- 770B
- Context window
- N/A
- Max output
- N/A
- Inputs
- text
- Outputs
- text
- Open weights
- Yes
- License
- Apache 2.0
Hy4 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionHy4 preview | Released | LLMBoard83.71 | Parameters770B | ContextN/A | Max outputN/A | Open weightsYes | LicenseApache 2.0 |
Models similar to Hy4
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
TE
Hy3
Tencent
OP
GPT-5.6-Terra
OpenAI
ZA
GLM 5.3 Flash
Zhipu AI
GO
Google
Gemini 3.7 Flash
AN
Claude Opus 4.7
Anthropic
ME
Muse Spark 1.1
Meta
What is Hy4?
Key information about Hy4 and its available data.
0 license. It supports software engineering, office and analysis tasks, game development, and scientific research, with a native MTP layer for speculative decoding.
Data as of 2026-09-08.
FAQ
Common questions about Hy4.
When was Hy4 released?
Hy4's default version was released on Aug 28, 2026.
How much does Hy4 cost?
No official standard PAYG price is currently available for Hy4. The lowest tracked third-party offer starts at $0.67 input and $2 output via Vancine.
Who created Hy4?
Hy4 was created by Tencent.
What is the context window for Hy4?
A context window is not available for the default version.
Is Hy4 open weight?
Yes. The default version is marked as open weight under Apache 2.0.
How many API providers offer Hy4?
9 provider offerings are linked to the default version.