Identity
- Status
- preview
- Developer ID
tencent/Hy4-preview- Released
- Aug 28, 2026
- Version
- Hy4 preview
- License
- apache-2.0
TencentHy4 preview
Hy4 preview is Tencent's open-weight flagship mixture-of-experts language model with 770B total parameters, 49B active parameters, and a 1M-token context window.
- Token context
- 1,000K tokens
- Inputs
- Text
- Outputs
- Text
- Released
- Aug 28, 2026
Providers & Pricing
Estimate workload cost →| Provider | Input / 1M | Output / 1M | Cache read / 1M | Cache write / 1M | Observed | Evidence |
|---|---|---|---|---|---|---|
 DeepinfraAvailable | $0.834 | $2.501 | $0.042 | — | Oct 19, 2026 | Provider-reported |
 OpenrouterAvailable | $0.834 | $2.501 | $0.042 | — | Oct 19, 2026 | Provider-reported |
Benchmarks
All benchmark results →| LMArena Agent Arenaagent-2026-10-16-a64ca3b12b6e | 2.68 | — | 84,828 | Hy4 preview; 95% CI [2.00929922, 3.35139547]; sessions 84828; observations 8314177; rank 20unknown | Original sourcewinner eligible | Third-party benchmark |
At a Glance
Model Facts
Capacity
- Token context
- 1,000K tokens
- Maximum output
- Unknown
- Total parameters
- 770B
- Active parameters
- 49B
- Knowledge cutoff
- Not reported
Interface and Access
- Inputs
- Text
- Outputs
- Text
- API available
- Yes
- Open weights
- Yes
- Self-hostable
- Yes
- Reasoning
- Yes
- Vision
- No
- Tool calling
- Yes
- Capabilities
- chat, generation, reasoning, tools
This page represents one developer model product. Dated API snapshots, serving endpoints, pricing, regions, quantizations, and service tiers attach as versioned aliases or provider details and never create duplicate public model records.
Version Lineage
PredecessorUnknown
SuccessorsUnknown
Aliaseshy4-preview, Hy4 preview, tencent/Hy4-preview, tencent/Hy4-preview-FP8
Explicit Unknowns
- Maximum output
- Tokenizer
- Knowledge cutoff
Primary Evidence
Sources and Observation Date
Comparable Peers
Related Model Comparisons
| A | Pair | B | Context |
|---|---|---|---|
Hy4 previewTencent | vs | Claude Fable 5.1Anthropic | cross-developer peerstext |
Hy4 previewTencent | vs |  Gemini 3.1 ProGoogle DeepMind | cross-developer peerstext |
Hy4 previewTencent | vs |  Kimi K3Moonshot AI | cross-developer peerstext |
Hy4 previewTencent | vs |  MiniMax M3MiniMax | cross-developer peerstext |
Hy4 previewTencent | vs |  GLM 5.3Z.ai | cross-developer peerstext |
Hy4 previewTencent | vs | cross-developer peerstext | |
Hy4 previewTencent | vs |  DeepSeek V4 ProDeepSeek | cross-developer peerstext |
Hy4 previewTencent | vs | cross-developer peerstext | |
Hy4 previewTencent | vs |  Solar Pro 4Upstage | cross-developer peerstext |
Hy4 previewTencent | vs |  Mistral Large 3Mistral AI | cross-developer peerstext |
Hy4 previewTencent | vs |  GPT-5.6 SolOpenAI | cross-developer peerstext |
Hy4 previewTencent | vs | Hy3Tencent | family variantstext |
Questions
Hy4 preview FAQs
What is Hy4 preview?+
Hy4 preview is Tencent's open-weight flagship mixture-of-experts language model with 770B total parameters, 49B active parameters, and a 1M-token context window. It is developed by Tencent and its current sourced lifecycle status is preview.
What inputs and outputs does Hy4 preview support?+
Hy4 preview's current record lists Text as input and Text as output.
How much context does Hy4 preview support?+
The current record lists a 1,000K-token context window and does not report a maximum output length.
Is Hy4 preview available through an API?+
Hy4 preview is marked as API-available. The directly sourced provider records currently include Deepinfra and Openrouter.
How large is Hy4 preview?+
The current record lists 770B total parameters and 49B active parameters. Its architecture is mixture_of_experts.
What capabilities and tasks does Hy4 preview support?+
Hy4 preview's recorded capabilities are chat, generation, reasoning, and tools. Its supported tasks are reasoning and text.
How much does Hy4 preview cost through an API?+
The lowest directly sourced prices currently attached to Hy4 preview are $0.834 per million input tokens through Deepinfra and $2.501 per million output tokens through Deepinfra. Prices are provider-specific and should be checked against each cited observation date.
Can Hy4 preview be self-hosted?+
Hy4 preview is an open-weight model. Self-hosting is supported, and the recorded license is apache-2.0.
Does Hy4 preview have related versions or aliases?+
No predecessor is recorded. No successor is recorded. Recorded aliases are hy4-preview, Hy4 preview, tencent/Hy4-preview, and tencent/Hy4-preview-FP8.
Which models can Hy4 preview be compared with?+
The current compatible peer set includes Claude Fable 5.1, Gemini 3.1 Pro, Kimi K3, MiniMax M3, and GLM 5.3. Compatibility requires shared sourced modalities and tasks; it does not imply that one model is better. Browse model comparisons →