Tencent · Benchmarks

Tencent Hy4 Preview benchmarks

Tencent Hy4 Preview has 3 published scores in this dataset, from a Tencent release dated 2026-08-28. Its composite index is 89.9, which places it 8th of 22 ranked models and 4th among 15 open-weight models. Its strongest area is reasoning (index 96.1) and its weakest is coding (index 83.7). Its best placing is on GPQA Diamond, where its 92.3 ranks 8th of 19 models with a published score.

Open weightsOfficial page

Category profile

Index 0–100: each score relative to the best published score on that test.

ReasoningCodingAgenticKnowledge
Reasoning
96.11 test · not ranked
Coding
83.7#6 of 13 · 2 tests
Agentic
—no data
Knowledge
—no data

Model facts

Lab
Tencent
Type
Open weights
Released
2026-08-28
Parameters
770B total, 49B active (MoE)
Context
1M
License
Apache 2.0

Tencent Hy4 Preview hardware requirements and cost

Even at 4-bit it needs about 531 GB (4× H200 (564 GB)), well beyond a home workstation, so most people run it on rented multi-GPU servers. Full BF16 precision needs about 1.8 TB (Multiple nodes (over 1.1 TB)).

Estimated GPU memory for Tencent Hy4 Preview
PrecisionWeightsWith headroomSmallest setup
BF16 (full precision)1.5 TB1.8 TBMultiple nodes (over 1.1 TB)
FP8 / 8-bit770 GB885 GB8× H200 (1,128 GB)
4-bit (GGUF Q4_K_M)462 GB531 GB4× H200 (564 GB)

Estimate: parameters × bytes per parameter, plus 15% for the KV cache and runtime at modest context lengths. Long contexts and large batches need more.

Lighter or stronger open models

Open models ranked just above and below on the composite index, with their 4-bit memory needs.

Tencent Hy4 Preview speed and pricing

Measured figures link to their source; estimates are labelled.

API output speed
52 tok/sOpenRouter median provider, last 30 min · OpenRouter model page provider stats
Time to first token
2.99 sOpenRouter model page provider stats
API price per 1M tokens
$0.834 in · $2.50 outlab's own endpoint on OpenRouter · OpenRouter endpoints API
Compare speed and price across all models

Score by test

Rank is among every model with a published score on that test.

GPQA Diamond92.3%Graduate-level, Google-proof science questions in biology, physics, and chemistry.#8 of 19 · 3.8 behind GPT-6 AstraSWE-bench Pro65.7%Harder, longer software tasks across many languages and repositories.#5 of 9 · 24.2 behind Claude Opus 5.5Terminal-Bench85.4%Terminal-Bench 2.1#6 of 13 · 5.2 behind DeepSeek V4.1 Flash

Nearby in the composite ranking

Models with the closest composite index.

  1. 1Claude Fable 5.191.2
  2. 2DeepSeek V4 Pro91.0
  3. 3Kimi K390.6
  4. 4Tencent Hy4 Preview89.9
  5. 5Muse Spark Open88.0
  6. 6Reflection Beam 1.087.0
  7. 7GPT-6.1 Sol86.5

Head-to-head comparisons

Side-by-side pages built from the tests both models report.

No head-to-head page for Tencent Hy4 Preview yet. Compare it on the rankings dashboard.

Sources

ModelTestScoreSettingSource
Tencent Hy4 PreviewGPQA Diamond92.3—Hugging Face model card (tencent/Hy4-preview), evaluation results
Tencent Hy4 PreviewSWE-bench Pro65.7—Hugging Face model card (tencent/Hy4-preview), evaluation results
Tencent Hy4 PreviewTerminal-Bench85.4Terminal-Bench 2.1Hugging Face model card (tencent/Hy4-preview), evaluation results

Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works