tencent
Tencent: Hy3 preview
tencent/hy3-preview
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
tencent/hy3-preview is the instruct variant described by Tencent as a 295B-parameter Mixture-of-Experts model with 21B active parameters and an added MTP layer. Its card reports improvements in complex reasoning, instruction following, context learning, coding and agent tasks, while OpenRouter lists text-to-text operation, a 262144-token context and configurable reasoning that defaults to high effort. Because the supplied evidence is vendor documentation plus a catalog entry, treat reported capability and performance claims as directional and validate them on the intended workloads.[1][2]
Strengths
- Sparse architecture: Tencent lists 295B total parameters, 21B active parameters, 192 experts with top-8 routing, and an MTP layer.[2]
- Long text context: both sources describe a 262144-token (256K) window with text input and output.[1][2]
- Operational controls: OpenRouter lists reasoning effort, tools, tool_choice, max_tokens and generation parameters; Tencent documents vLLM and SGLang deployment.[1][2]
Limitations
- No independent evaluation is supplied; benchmark evidence is vendor-reported, partly image-based, and its numeric comparison table concerns Hy3 preview-Base rather than this exact instruct ID.[2]
- Reasoning defaults conflict: OpenRouter lists high effort, while Tencent's quickstart calls no_think the default.[1][2]
- The card names the Tencent Hy Community License Agreement and an eight-GPU serving path, but the supplied evidence lacks full license terms and broader hardware sizing.[2]
Where to use it
- Evaluate agentic coding or search-agent workflows that can use tool controls and adjustable reasoning; reported benchmark fit remains workload-dependent.[1][2]
- Consider long-document analysis, context learning, or complex instruction following where text-only input and output are sufficient.[1][2]
- Consider a self-hosted pilot through the documented vLLM or SGLang paths if BF16 precision and the reported multi-GPU setup fit the environment.[2]
Sources
- [1] OpenRouter catalog · Tencent: Hy3 previewSources checked: Nov 13, 2026
- [2] Model documentation · Tencent: Hy3 previewSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Hy3 preview is a high-efficiency Mixture-of-Experts model from Tencent designed for agentic workflows and production use. It supports configurable reasoning levels across disabled, low, and high modes, allowing it to...
Specifications
- Model ID
- tencent/hy3-preview
- Canonical ID
- tencent/hy3-preview-20260421
- Creator
- tencent
- Added to OpenRouter
- 22.04.2026
- Context length
- 262,144
- Modality
- text->text
- Input modalities
- text
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.180 / out $0.600
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | reasoning: optional · default high · efforts: high, low, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Coding | BB — next 30% of the cohort | — |
| Agents | BB — next 30% of the cohort | parameters: tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · default high · efforts: high, low, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Coding index | 58.8 | BB — next 30% of the cohort | 13.11.2026 |
| Artificial Analysis | Agentic index | 24.1 | BB — next 30% of the cohort | 13.11.2026 |
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| GMICloudbf16 | 262,144 | $0.180 | $0.600 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.