Tencent Hy3
Tencent's open-weights agent model (April 2026), consistently top-3 on OpenRouter token-usage rankings. A 295B MoE with only 21B active parameters, it sustains agentic tool-use chains up to ~495 steps with configurable reasoning levels at rock-bottom prices. Served as 'hy3-preview' on OpenRouter, but it is GA (the name is a naming choice, not an access gate); also on Tencent Cloud TokenHub.
Tencent Hy3 strengths
- Top-3 OpenRouter usage
- ~495-step agentic tool chains
- Very low cost (21B active MoE)
- Configurable reasoning levels
- Open weights, self-hostable
Pricing & context
| Context window | 256K tokens |
| Input price /1M | $0.063 |
| Output price /1M | $0.21 |
| Modalities | text |
Cost guide: a typical call of about 10K input + 2K output tokens costs roughly $0.001 at list prices. Worth modelling against cheaper tiers before committing high-volume traffic.
When to choose Tencent Hy3
Tencent Hy3 is best for cost-efficient agent pipelines, coding agents and high-volume production workloads. If your workload is more cost-sensitive, weigh it against gpt-oss-120b (≈$0.03 input /1M) first.
Tencent Hy3 FAQ
How much does Tencent Hy3 cost?
Tencent Hy3 is priced at $0.063 per 1M input tokens and $0.21 per 1M output tokens (public API list price), with a 256K tokens context window. A typical call of about 10K input and 2K output tokens costs roughly $0.001.
What is Tencent Hy3 best for?
Tencent Hy3 by Tencent (Hunyuan) is best for cost-efficient agent pipelines, coding agents and high-volume production workloads.
How does Tencent Hy3 pricing compare to Qwen3.7 Max?
Tencent Hy3 input costs $0.063 per 1M tokens versus $1.25 for Qwen3.7 Max, roughly 19.8x less expensive on input. Output is $0.21 vs $3.75.
Is Tencent Hy3 multimodal?
Tencent Hy3 supports text.
Other models
All models →| 01 | Claude Fable 5 | Anthropic | $10.00 | → |
| 02 | GPT-5.5 | OpenAI | $5.00 | → |
| 03 | Claude Opus 4.8 | Anthropic | $5.00 | → |
| 04 | Gemini 3.1 Pro | $2.00 (under 200K; $4.00 above) | → |