DeepSeek: DeepSeek V3.2 vs Tencent: Hy3
Based on weekly OpenRouter usage tracked by AI Tier List, Tencent: Hy3 ranks #2 overall — ahead of DeepSeek: DeepSeek V3.2 at #28. Tencent: Hy3 is cheaper on input tokens at $0.13/1M vs $0.27/1M for DeepSeek: DeepSeek V3.2. Tencent: Hy3 offers the larger context window at 262K tokens.
DeepSeek: DeepSeek V3.2 vs Tencent: Hy3: which should you use?
Choose DeepSeek: DeepSeek V3.2 if…
- Cheaper to write — $0.40 vs $0.53 per 1M output tokens, the line item that decides generation-heavy costs.
Choose Tencent: Hy3 if…
- It sees more real traffic — #2 overall on OpenRouter's weekly ranking, against #28 for DeepSeek: DeepSeek V3.2.
- Cheaper to read — $0.13 vs $0.27 per 1M input tokens, which is what dominates long-prompt and RAG workloads.
- Bigger context window — 262K vs 164K tokens.
- It is the newer model — released Jul 2026 against Dec 2025.
Spec Comparison
| Spec | DeepSeek: DeepSeek V3.2 | Tencent: Hy3 |
|---|---|---|
| Author | deepseek | tencent |
| Input price (per 1M tokens) | $0.27 | $0.13 |
| Output price (per 1M tokens) | $0.40 | $0.53 |
| Context window | 164K | 262K |
| Input modalities | text | text |
| Output modalities | text | text |
| AA Intelligence | 32.6 | — |
| AA Coding | 44.2 | — |
| AA Agentic | 18.3 | — |
| Released | Dec 2025 | Jul 2026 |
What DeepSeek: DeepSeek V3.2 and Tencent: Hy3 actually cost to run
On a read-heavy job — 1M tokens in, 100K out — DeepSeek: DeepSeek V3.2 costs $0.31 and Tencent: Hy3 costs $0.18.
Flip the shape — 100K in, 1M out — and it becomes $0.43 for DeepSeek: DeepSeek V3.2 versus $0.54 for Tencent: Hy3.
Both figures are the published per-token rates multiplied out; caching and batch discounts are not included.
Usage Rank by Task
| Task | DeepSeek: DeepSeek V3.2 | Tencent: Hy3 |
|---|---|---|
| Overall | #28 | #2 |
| Memory Extraction | — | #3 |
| Multi-step Planning | — | #3 |
| Tool Dispatch | — | #3 |
| Web Search | — | #3 |
| Agent Workflows | — | #1 |
| Classification & Tagging | — | #10 |
| Debugging | — | #9 |
| DevOps Config | — | #7 |
| File Read/Write | — | #2 |
| Frontend UI | — | #6 |
| Code Implementation | — | #4 |
| Repo Analysis | — | #2 |
| Code Review & Security | — | #8 |
| Shell Execution | — | #4 |
About DeepSeek: DeepSeek V3.2 and Tencent: Hy3
DeepSeek: DeepSeek V3.2
Built by deepseek · released Dec 2025 · 164K-token context
DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
Tencent: Hy3
Built by tencent · released Jul 2026 · 262K-token context
Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use. It supports a configurable reasoning effort:...
Related comparisons
- DeepSeek: DeepSeek V3.2 vs DeepSeek: DeepSeek V4 Flash 0731
- DeepSeek: DeepSeek V3.2 vs OpenAI: GPT-5.6 Luna (batch)
- DeepSeek: DeepSeek V3.2 vs DeepSeek: DeepSeek V4 Flash 0423
- Tencent: Hy3 vs DeepSeek: DeepSeek V4 Flash 0731
- Tencent: Hy3 vs OpenAI: GPT-5.6 Luna (batch)
- Tencent: Hy3 vs DeepSeek: DeepSeek V4 Flash 0423
Get notified when DeepSeek: DeepSeek V3.2 or Tencent: Hy3 moves
Usage ranks and prices refresh weekly. We email you when they move — once a week, no spam.
FAQ
Which is better, DeepSeek: DeepSeek V3.2 or Tencent: Hy3?
Based on weekly OpenRouter usage tracked by AI Tier List, Tencent: Hy3 ranks #2 overall — ahead of DeepSeek: DeepSeek V3.2 at #28. Tencent: Hy3 is cheaper on input tokens at $0.13/1M vs $0.27/1M for DeepSeek: DeepSeek V3.2. Tencent: Hy3 offers the larger context window at 262K tokens.
Which is cheaper, DeepSeek: DeepSeek V3.2 or Tencent: Hy3?
Tencent: Hy3 is cheaper at $0.13 per 1M input tokens. Output pricing: DeepSeek: DeepSeek V3.2 $0.40, Tencent: Hy3 $0.53.
Is Tencent: Hy3 still ahead of DeepSeek: DeepSeek V3.2 right now?
Yes. For the week of 2026-08-10, Tencent: Hy3 ranks #2 overall on OpenRouter usage against #28 for DeepSeek: DeepSeek V3.2. Ranks refresh weekly, and the verdict on this page flips if they cross.
Usage ranks are weekly data aggregated by OpenRouter. Prices are USD per 1M tokens.