Hy3 vs Hy3 preview
Benchmarks, API Token Pricing, Context Window & Speed Comparison
Related topics (12)
Tencent 的 Hy3 与 Tencent 的 Hy3 preview 是当前备受关注的两款主力 AI 模型。Hy3 提供 262.1K Token 上下文窗口，API 定价为每百万 Token 输入 $0.00 / 输出 $0.00；而 Hy3 preview 提供 262.1K Token 上下文窗口，定价为每百万 Token 输入 $0.00 / 输出 $0.00。以下为两者的基准测试、延迟与成本全面对比。
切换并对比任意模型
Select any AI models from our verified benchmark database to generate live side-by-side matrices.
Hy3
开源权重推理大模型Tencent 推出的 Hy3 是一款通过大规模强化学习训练、支持私有化部署的开源权重推理大模型，具备 262K 上下文窗口、85 Token/秒生成速度与 81/100 综合智能指数（免费开源权重）。
$0.00 / 1M
$0.00 / 1M
262.1K tokens
N/A
Hy3 preview
开源权重基础大模型Tencent 推出的 Hy3 preview 是一款支持私有云服务器自托管、微调与完全数据主权的开源权重基础大模型，具备 262K 上下文窗口、88 Token/秒生成速度与 79/100 综合智能指数（免费开源权重）。
$0.00 / 1M
$0.00 / 1M
262.1K tokens
N/A
核心差异一览
- Both models cost the same under this workload — choose on capability, not price.
- Hy3 preview is faster at 88 tok/s vs 85 tok/s.
| Feature | Hy3 | Hy3 preview |
|---|---|---|
Intelligence Index Composite reasoning & knowledge score (0-100) | 81 | 79 |
Provider | Tencent | Tencent |
Input Price Cost per million input tokens | $0.00 / 1M | $0.00 / 1M |
Output Price Cost per million output tokens | $0.00 / 1M | $0.00 / 1M |
Blended Price Weighted 7:2:1 cache:input:output per 1M | $0.00 / 1M | $0.00 / 1M |
Context Window Maximum input + output tokens | 262.1K tokens~393 A4 pages of text | 262.1K tokens~393 A4 pages of text |
Max Output Tokens Per single request | 16K | 8K |
Throughput Speed Tokens generated per second | 85 tok/s | 88 tok/s |
SWE-bench Verified Real-world software engineering | N/A | N/A |
GPQA Diamond Graduate-level reasoning | 64% | N/A |
各项指标谁更胜一筹？
At-a-glance decision frameworkEvaluations ongoing
权威基准测试得分对比
Official evaluation metrics verified by independent third-party research labs.
Challenging questions created by domain experts to test complex PhD-level reasoning.
Hardened reasoning and academic knowledge benchmark with reduced lucky-guess probability.
Crowdsourced human preference evaluation based on blind, head-to-head model battles.
Tokens emitted per second during active generation streaming.
延迟与响应速度
Time to first token (TTFT) and estimated end-to-end response time for a typical 500-token response.
实时 Token 成本模拟器
Adjust sliders to calculate projected monthly cloud API spending with prompt caching discounts.
Hy3 · Strengths & Key Advantages
- 提供 262K 上下文窗口与最高 16K 输出容量，核心基准表现优异（GPQA Diamond 64% · MMLU-Pro 79%）
- 高达 85 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 460 毫秒
- 开放模型权重 (Confidential / 140B MoE)，支持通过 vLLM、SGLang 或 Ollama 在私有 GPU 集群上本地部署与微调
Considerations & Trade-offs
- 在本地以满血精度运行需要较高的 GPU 显存 (VRAM) 与运维基础设施
- 在超长链路的前沿科研推导任务上，深度略逊于顶配重型旗舰推理模型
Hy3 preview · Strengths & Key Advantages
- 提供 262K 上下文窗口与最高 8K 输出容量，核心基准表现优异（MMLU-Pro 76.5% · Arena Elo 1310）
- 高达 88 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 440 毫秒
- 开放模型权重 (Confidential / 140B MoE)，支持通过 vLLM、SGLang 或 Ollama 在私有 GPU 集群上本地部署与微调
Considerations & Trade-offs
- 在本地以满血精度运行需要较高的 GPU 显存 (VRAM) 与运维基础设施
- 在超长链路的前沿科研推导任务上，深度略逊于顶配重型旗舰推理模型
Token Use & Task Economics
Estimated cost and token consumption to complete one representative agentic task.
| Metric | Hy3 | Hy3 preview |
|---|---|---|
Cost per Task Estimated USD to finish one task | $0.00 | $0.00 |
Output Tokens / Task Tokens generated per task | 18K | 15K |
Reasoning Tokens / Task Thinking tokens burned per task | 5K | 3K |
技术规格参数
Architecture, release timeline, capabilities, and modalities.
| Feature | Hy3 | Hy3 preview |
|---|---|---|
| Official API Model ID (SDK) | hy3 | hy3-preview |
| API Protocol | OpenAI-Compatible REST / vLLM / SGLang / HuggingFace TGI | OpenAI-Compatible REST / vLLM / SGLang / HuggingFace TGI |
| Release Date | 2026-07 | 2026-04 |
| Knowledge Cutoff | April 2026 | February 2026 |
| Architecture | Efficient Open Reasoning MoE Transformer | Preview Open Reasoning MoE Transformer |
| Parameters | Confidential / 140B MoE | Confidential / 140B MoE |
| Enterprise Rate Limits (RPM / TPM) | Unlimited (Hardware-bound on self-hosted GPU) / 10,000 RPM Cloud · Unlimited local throughput / 10M+ TPM Managed Cloud | Unlimited (Hardware-bound on self-hosted GPU) / 10,000 RPM Cloud · Unlimited local throughput / 10M+ TPM Managed Cloud |
| Long-Context Tier (>128K/200K) | Free across full context window (only KV-cache VRAM scales) | Free across full context window (only KV-cache VRAM scales) |
| Prompt Cache Policy & TTL | Automatic prefix KV-caching in vLLM / SGLang (0 cost) | Automatic prefix KV-caching in vLLM / SGLang (0 cost) |
| Computer Use / GUI Agent | API / CLI Tool-Use & DOM Agent ready (no raw pixel desktop click loop) | API / CLI Tool-Use & DOM Agent ready (no raw pixel desktop click loop) |
| MCP & Parallel Tool Calling | Native MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema | Native MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema |
| Web Search Grounding & Sandbox | Supported via external Tavily/Exa Search & E2B/Docker Sandbox tools | Supported via external Tavily/Exa Search & E2B/Docker Sandbox tools |
| Official Fine-Tuning Support | Full Weights SFT, LoRA, QLoRA, DPO & GRPO RL (Unsloth / Axolotl / Llama-Factory) | Full Weights SFT, LoRA, QLoRA, DPO & GRPO RL (Unsloth / Axolotl / Llama-Factory) |
| Multimodal Token Billing | Text BPE (~4 chars/token) | Text BPE (~4 chars/token) |
| Quantization & Min GPU VRAM | BF16 / FP16, FP8 Dynamic, INT4 AWQ / GPTQ, GGUF (Q4_K_M / Q8_0) · 87 GB VRAM (4-bit Quantized) · 301 GB VRAM (Native BF16) | BF16 / FP16, FP8 Dynamic, INT4 AWQ / GPTQ, GGUF (Q4_K_M / Q8_0) · 87 GB VRAM (4-bit Quantized) · 301 GB VRAM (Native BF16) |
| Safety & Frontier Classification | Open-Weights System Card + Customizable Guardrail Layer (Llama Guard / ShieldGemma) | Open-Weights System Card + Customizable Guardrail Layer (Llama Guard / ShieldGemma) |
| Enterprise Compliance (ZDR/HIPAA/SOC2) | 100% Air-Gapped On-Premises / Private VPC · Full HIPAA, GDPR & SOC2 Control | 100% Air-Gapped On-Premises / Private VPC · Full HIPAA, GDPR & SOC2 Control |
| Input Modalities | text | text |
| Output Modalities | text | text |
Hy3 vs Hy3 preview: 常见问题解答 (FAQ)
Which is more intelligent, Hy3 or Hy3 preview?
Hy3 is more intelligent. It scores 81 compared with 79 on the composite Intelligence Index.
Which model is faster?
Hy3 preview is faster, generating 88 tokens per second compared with 85 tokens per second.
Which model has lower latency?
Hy3 preview has lower latency, with a time to first token of 0.44s compared with 0.46s.
综合结论
综合来看，Hy3（适合：通过大规模强化学习训练、支持私有化部署的开源权重推理任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。）与 Hy3 preview（适合：支持私有云服务器自托管、微调与完全数据主权的开源权重基础任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。）各具优势：Hy3 在综合智能与编程基准测试中表现更强，Hy3 preview 拥有更快的 Token 生成速度，而 Hy3 则在规模化 API 部署中具备更高的成本效益。
常见问题 (FAQ)
在编程和深度推理方面，Hy3 和 Hy3 preview 哪个更强？+
在综合智能指数与编程基准评测中，Hy3 表现占优。建议结合上方的 SWE-bench Verified 与 GPQA Diamond 跑分根据具体场景选择。
Hy3 与 Hy3 preview 的 API 价格哪个更便宜？+
Hy3 的 API 价格为输入 $0.00 / 输出 $0.00（每百万 Token），Hy3 preview 为输入 $0.00 / 输出 $0.00。综合混合成本来看，Hy3 更具性价比。
哪款模型拥有更大的上下文窗口和更快的响应速度？+
Hy3 支持 262.1K Token 上下文窗口，而 Hy3 preview 支持 262.1K Token。在流式生成速度方面，Hy3 preview 的吞吐表现更快。
Understand key metrics like context window, token pricing, reasoning mode, and SWE-bench scores in our comprehensive AI glossary.
Related reading
- Gemma 4 E4B vs Hy3 preview：基准测试、API 价格与速度全面对比 (2026)对比
- Hy3 vs Kimi K2.7 Code：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-3.0 vs Aion-Mini：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-3.0 vs Gemini 3.6 Flash：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-Mini vs Laguna XS 2.1：基准测试、API 价格与速度全面对比 (2026)对比
- Aya Expanse 32B vs Llama 3.2 90B Vision：基准测试、API 价格与速度全面对比 (2026)对比
Benchmark scores, token prices, and rate limits evolve rapidly as AI labs release model updates. Always verify current specifications and pricing on the official provider documentation.
数据来源: Tencent 与 Tencent 官方 API 文档、模型发布说明及独立基准测试报告 (2026)。