Skip to content
Bencher

Models · Tencent Hunyuan · Out sinceReleased 28 Aug 2026

Hy4 preview

Available on OpenRouterOn OpenRouterAvailable nowGenerally availableCan run on your own serversOpen weights

Hy4 preview is made by Tencent Hunyuan. We don't have enough test results yet to rank it. It's cheap to use. Its makers have published it, so you can run it on your own servers.

Auto-created from OpenRouter catalog; verify details.

Writing & creativity
—
Research & analysis
—
Coding
—
Price
Cheap$0.83 / $2.50
Price per 1M (blended)Blended / 1M
$1.25
MemoryContext
1M
Longest answerMax output
—
Test resultsResults
7 (7 independent7 indep.)
Out sinceReleased
28 Aug 2026
Made byVendor
Tencent Hunyuan
UnderstandsInputs
—

Scores are out of 100 for each area, compared with every model we track; “#” is its rank. Context is how much text it can read at once (1M is roughly 700,000 words). Blended price mixes the cost of what you send and what it writes back.

Thinking levels

Reasoning settings

You can't choose how long this one thinks.

No adjustable reasoning setting listed.

Read more

Links

Using it through OpenRouter

On OpenRouter

OpenRouter is a service that gives access to many AI models in one place. This model is listed there as tencent/hy4-preview.

Route it as tencent/hy4-preview at $0.83 in / $2.50 out per 1M tokens, 1M context. Listed since 28 Aug 2026.

frequency_penaltyinclude_reasoninglogit_biasmax_completion_tokensmax_tokensmin_ppresence_penaltyreasoningreasoning_effortrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_p

Run it yourself

Self-hosting

On your own servers,Licence, sizeunder your own control.and hardware.

Compare all open models →All open-weights models →

LicenceLicence
Apache-2.0
OK for business useCommercial use
Yes
SizeParams (total / active)
770B / 49B
Can be trained furtherBase model
No

Hardware you'd need

Hardware tier & memory

Needs one full AI server.

≤ 1.1 TB at 8-bit (one 8×H200 node). Weights ≈ 809 GB at 8-bit, 424 GB at 4-bit (+10–30% for KV cache). MoE: 49B active per token.

Languages: Languages: Not stated.

Training it further: Fine-tuning: Official finetuning pipeline in repo (finetune/README.md); AngelSlim toolkit for compression.

Quantisations: bf16, fp8 (tencent/Hy4-preview-FP8). Engines: vLLM, SGLang.

Download (Hugging Face)Weights on Hugging Face ↗huggingface.co ↗huggingface.co ↗

Every test result

Every result

All the numbers,with where they came from.

Every result we've found for this model, with who measured it and a link to where we read it.

All raw rows (vendor and independent kept separate), with setting label, source, date, harness and notes.

All results for this model
NotesNotes
IOI (Vals)59.3%DefaultIndependent testIndependentVals.ai ↗29 Sep 2026—±4.596 stderr; $1.355/test
LMArena Code Arena (WebDev)1633Defaulthy4-previewIndependent testIndependentLMArena ↗30 Sep 2026—Code Arena | WebDev overall (agentic web-dev, raw); rank 16 (CI rank 14-23); 95% CI 1624-1643; 5088 votes
ProgramBench (fully resolved)0.0%DefaultIndependent testIndependentVals.ai ↗29 Sep 2026mini-SWE-agent±0 stderr; $12.809/test; strict fully-resolved rate
Vals Code Migration47.4%DefaultIndependent testIndependentVals.ai ↗29 Sep 2026—±4.272 stderr; $3.405/test
Vals Legal Research Bench45.2%Defaultvals id tencent/hy4-previewIndependent testIndependentVals.ai ↗29 Sep 2026—vals id tencent/hy4-preview; rank 15/72; ±3.459 stderr; $0.846372/test
Vals Public Benefits Bench v1.168.6%Defaultvals id tencent/hy4-previewIndependent testIndependentVals.ai ↗29 Sep 2026—vals id tencent/hy4-preview; rank 7/45; ±1.207 stderr; $0.252532/test
Vals SRE Bench2.3%DefaultIndependent testIndependentVals.ai ↗29 Sep 2026—±0.926 stderr; $4.902/test