Hy3 by Tencent (via OpenRouter)
No longer free
Hy3 is Tencent's 295B MoE open-weight model (21B active). OpenRouter no longer serves tencent/hy3:free for free (checked Nov 3, 2026). Requests return a 404 that points to the paid ID tencent/hy3, which had 5 endpoints, from $0.132 per 1M input tokens and $0.528 per 1M output tokens.
See which OpenRouter free models still work, or browse the models that are free today.
Benchmarks
Numbers are copied from the linked public leaderboards for this exact model and version. The free endpoint can run a different reasoning setting or quantization than the tested one.
Replacing a retired model in production?
We wire free-tier and open-weight models into production stacks: evals against your real inputs, failover, cost caps. Book a free 30-minute call.
Frequently asked
Is Hy3 still free?
No. Hy3 is Tencent's 295B MoE open-weight model (21B active). OpenRouter no longer serves tencent/hy3:free for free (checked Nov 3, 2026). Requests return a 404 that points to the paid ID tencent/hy3, which had 5 endpoints, from $0.132 per 1M input tokens and $0.528 per 1M output tokens.
What should I use instead of tencent/hy3:free?
Use the paid ID if the price fits, or move to a model that is still free and keep a paid fallback behind it. Our check of every OpenRouter :free ID lists what still answers today and what each dead ID costs on its paid slug: https://www.ayautomate.com/blog/openrouter-free-models
Free monthly update
The monthly free-models update
What is newly free, what got rate-limited, and what to switch to. One email a month. Unsubscribe anytime.
Related free models
- GLM 4.6 (Z.ai) · Self-host free
- DeepSeek V3-0324 (DeepSeek) · Self-host free
- Gemini 2.5 Pro (Google Gemini) · Free tier · existing users only, shuts down Dec 1, 2026
- Qwen3.7 Max (Qwen) · One-time 1M tokens, 90 days (new users)
- Gemini 2.5 Flash (Google AI Studio) · Free tier · existing users only, shuts down Dec 1, 2026
- + 90 more models (NVIDIA NIM) · Prototyping only · rate limits not published