Hy4-preview: benchmark scores, price and ranking

Hy4-preview from tencent ranks #91 of 743 LLMs on the LLMs Tiger aggregate benchmark score (51.9), based on 2 benchmark results captured up to 2026-10-13.

Organisation: tencent
Type: Open-weight / local model
Open weights: Yes
Cheapest API price: $0.751 in / $2.25 out per million tokens (via openrouter, observed 2026-10-13T20:47Z)
Aggregate rank: #91 of 743 (score 51.9)

Across the 2 benchmarks where at least ten models are scored, Hy4-preview is in the top 10% on 1 and in the top half on 2.

Its cheapest listed input price, $0.751 per million tokens, is lower than that of 40% of the 427 models with a listed price.

Hy4-preview benchmark results

Hy4-preview benchmark scores, ranked against all models measured on each benchmark
Benchmark	Score	Rank	Source	Captured
GPQA Diamond	92.3%	#20 of 275	HuggingFace GPQA Diamond leaderboard	2026-10-13
LMArena Code Elo	1,632	#12 of 30	LMArena	2026-10-13

About Hy4-preview

Hy4-preview is an open-weight model that can run on your own hardware from tencent. The cheapest API listing we track charges $0.751 per million input tokens and $2.25 per million output tokens, through openrouter. We hold 2 benchmark results for it across 2 areas, and 1 is in the top tenth of its board.

Where Hy4-preview stands

On our aggregate score, Hy4-preview is #91 of 743 ranked models, in the top quarter. Its score is 51.9; the scale is centred on 50, so it sits 1.9 points above the typical model. That beats 88% of the ranked field. Among the 5 ranked tencent models it is #2. The next one up is Hy3 at 51.9. The next one down is Hy3-preview at 51.6.

Hy4-preview, benchmark by benchmark

On GPQA Diamond, Hy4-preview scores 92.3% and ranks #20 of 275. Only 19 models score higher, led by GPT-6 Astra (max) at 95.8%, which is 3.5 points ahead. It covers expert-level science questions. The board median is 71.7%, so Hy4-preview is 20.6 points above it. Source: HuggingFace GPQA Diamond leaderboard, captured 2026-10-13.

On LMArena Code Elo, Hy4-preview reaches 1,632. It ranks #12 of 30, ahead of 60% of the field, and 182 points behind Claude Opus 5.5 (max) at 1,814. The results either side are MiMo-V2.6 Flash (1,636) and Qwen3.8-Flash-Next (1,632). It covers crowd-voted coding answers. The board median is 1,622, so Hy4-preview is 10 points above it. Source: LMArena, captured 2026-10-13.

Strengths and weaknesses

Hy4-preview has too few results on boards with ten or more models to say where it is strong or weak. Treat any single score as a data point, not a verdict.

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so Hy4-preview costs more than the typical listing. Models that score within 1.5 points or better and cost less: DeepSeek V4 Flash 0731 (max) ($0.016, 53.1), Qwen3.7 Flash ($0.030, 50.5), GLM-5.2 (high) ($0.060, 51.6). Similar scores cost more on claude-3-5-haiku-20241022 ($1.00) and nousresearch/hermes-3-llama-3.1-405b ($1.00).

Reading these numbers

The 2 scores on this page come from 2 sources: HuggingFace GPQA Diamond leaderboard (1), LMArena (1). No official leaderboard supplied any of them, so treat them as third-party measurements. With 2 results the aggregate rank can move a lot when one new score arrives.

Models ranked nearby

#88 Hy3 (51.9)
#89 Dots3-Note Preview (Max) (51.9)
#90 GPT-5.2 (xhigh) (51.9)
#92 gpt-4o-2024-05-13 (51.9)
#93 ollama/qwen2.5-coder:32b (51.9)
#94 Claude 4.7 (Max) (51.8)