Hy3 on QuickSilver Pro
Hy3 is Tencent's Apache-2.0 open-weight 295B Mixture-of-Experts model (21B active) with a 262K-token context. QuickSilver Pro pins Tencent's official route and charges $0.091 input / $0.363 output per 1M tokens, within about 10% of Tencent's own $0.0825 / $0.33 rate.
At a glance
Tencent's Apache-2.0 295B MoE (21B active) for general-purpose coding and agentic workflows over a 262K-token context.
Pricing comparison ($/1M tokens)
| Provider | Input | Output | vs QSP |
|---|---|---|---|
| QuickSilver Pro | $0.091 | $0.363 | — |
| OpenRouter (tencent/hy3) | $0.0825 | $0.33 | 10% more expensive |
When to use
Reach for Hy3 on general-purpose coding and agentic workflows that fit a 262K-token context: multi-file edits, tool-using agents, and everyday code generation. Tencent ships it Apache-2.0 as a 295B Mixture-of-Experts model that activates 21B parameters per token, so you get a large-model knowledge base at a mid-tier per-token price. Tencent reports ~78 on SWE-Bench Verified.
When to use something else
For tasks that need more than 262K tokens of context, use a 1M-context model such as Qwen 3.7 Flash, MiniMax M3, or Kimi K3. For the lowest token price, Qwen 3.7 Flash costs less; for frontier reasoning, use DeepSeek V4 Pro or Kimi K3.
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/chat/completions \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "hy3",
"messages": [{"role": "user", "content": "Hello!"}]
}'OpenAI-compatible. One-line migration via base_url.
FAQ
Tencent describes Hy3 as an Apache-2.0 open-weight 295B-parameter Mixture-of-Experts model that activates 21B parameters per token, with a 262K-token context window. QuickSilver Pro serves it in non-thinking mode, so replies are direct with no reasoning trace.
Yes — Hy3 is an OpenAI-compatible chat completions endpoint on QuickSilver Pro. Set base_url=https://api.quicksilverpro.io/v1, paste your QSP key, and use model="hy3". Streaming, tool calling, json_schema strict mode, and usage.cost accounting all work. Migration is a base_url + key swap, dropping the `tencent/` provider prefix from the model ID.
QuickSilver Pro lists Hy3 at $0.091 input / $0.363 output per 1M tokens, about 10% above Tencent's own $0.0825 / $0.33 rate on OpenRouter. QSP pins Tencent's official endpoint for consistent reasoning behavior; drop the `tencent/` prefix when migrating.