Hy4 preview
tencent/hy4-preview- Tencent
- Family · Hy4
- Open weights
- Text → Text
- Reasoning · adjustable
- Tools · JSON
Not on OSMEXO yet
Spec & list price
Context1M
Max output64K
Input / 1M$0.83
Output / 1M$2.503× input
Usage
Tokens / day734.6B▲ 0.4% 7d
Usage rank#7
Requests / day6.5M
Tool calls / req0.92
Per request · 30 days
Avg input105.5K7.1× median
Avg output1.0K1.5× median
Cache hit95%1.4× median
Reasoning share53%about median
Tool-call failures0.2%0.1× median
Consumption
Peak 1.7T 29 OctLow 665.2B 2 OctRank today #7
News & trends
Trending · 14 days
News
- sglang v0.5.20SGLang · 18 Sep
- transformers v5.17.0: Release 5.17.09 Sep
- vllm v0.29.0vLLM · 9 Sep
Speed
Output speed#246/33446 tok/s1.4× slower than median
Fastest provider81 tok/sbest of 7 providers
First token#2842.78 s2.7× longer than median
Technical details
- Reasoning
- on · high
- Tokenizer
- Other
- License
- Apache 2.0
- Parameters
- 780B
- Moderation
- none
- Downloads · likes
- 11.4K · 523
Reasoning effort
- high · default
- low
- none
Request parameters 22
- tools
- tool_choice
- parallel_tool_calls
- response_format
- structured_outputs
- reasoning
- reasoning_effort
- include_reasoning
- temperature
- top_p
+12 more
- top_k
- min_p
- max_tokens
- max_completion_tokens
- stop
- seed
- frequency_penalty
- logit_bias
- logprobs
- presence_penalty
- repetition_penalty
- top_logprobs
When to use
Best for
- Very long document analysis
- Tool-calling and structured-output pipelines
- Self-hosted experiments with a newer open model
- Reasoning toggled on or off per request
Think twice for
- Production use needing a stable release
- Image or audio input
- Work needing published benchmark evidence
Research papers
Want Hy4 preview on OSMEXO?
Sign in to request it. Requests are reviewed; a request does not guarantee availability.
