Tencent / tencent/hy3

Tencent: Hy3 - access through LLMTR

Tencent Hy3 is a 295B-parameter Mixture-of-Experts model that activates roughly 21B parameters per token, built for coding agents, long-document analysis, multi-turn dialogue and multi-step tool workflows. Thinking is off by default; the `:low` and `:high` suffixes turn it on shallow or deep, and the reasoning chain comes back separately in `reasoning_content`, with its tokens counted inside `completion_tokens`. It supports native tool calling, schema-enforced JSON output through `json_schema`, and prompt caching on repeated prefixes. Use `tool_choice` with `auto` or `required`; forcing one named function is not reliable on this provider. The model is advertised with a 256K window, but a single request accepts at most 192K tokens (196,608) of input: beyond that the provider silently drops the OLDEST part of the prompt and answers from what is left, without reporting the loss. LLMTR therefore publishes 192K as this model's context length, so keep long documents under it or chunk them. Text only, no image or audio input.

Technical specifications

Canonical IDtencent/hy3
ProviderTencent
Context window196,608 tokens
OperationsCHAT_COMPLETIONS
Modalitiestext

Pricing

An 8% platform margin applies to credit top-ups; model usage prices are not separately marked up.

OperationMetricUnitPrice
CHAT_COMPLETIONSINPUT_TEXTPER_1M_TOKENS$0.140000
CHAT_COMPLETIONSCACHE_READPER_1M_TOKENS$0.035000
CHAT_COMPLETIONSOUTPUT_TEXTPER_1M_TOKENS$0.580000

Example usage

With existing OpenAI SDK flows, change only the base URL and model identifier.

curl https://llmtr.com/v1/chat/completions \
  -H "Authorization: Bearer llmtr-your_key" \
  -H "Content-Type: application/json" \
  -d '{"model":"tencent/hy3","messages":[{"role":"user","content":"Hello"}]}'

Related models