Compare Models

Hy4 Preview

Hy4 Preview is an open-weight model from Tencent, released August 28, 2026. It ranks #22 of 32 models on the Vals Index with 53.21%. Its best result is #9 of 50 on Public Benefits Bench v1.1.

Release Date: Aug 28, 2026

Developer Tencent  🇨🇳
Context Window 1M
Max Output Tokens 64k
Token Costs (in/out) $0.83/2.50
Weights Open
Input Modalities

Accuracy

53.21 % ± 1.13

Cost / Test (Vals Index)

$ 1.568

Latency

59 min 23 s

Vals Index
BenchmarksAccuracyRankings
Vals Index

53.21%

±1.13
22/32
Vals Finance Index

57.51%

±1.45
14/35
Vals Legal Index

71.70%

±1.84
12/33
Code Migration v1.1

45.36%

±3.59
19/29
CyberBench v1.1

65.36%

±5.55
32/47
EMB v1.1

68.95%

±2.40
21/37
Finance Agent v2.1

61.62%

±2.06
16/35
Legal Research Bench

45.19%

±3.46
21/89
MedCode

43.25%

±2.13
47/107
MedScribe

83.60%

±2.06
45/109
ProofBench v1.1

75.00%

±4.35
17/52
Public Benefits Bench v1.1

68.61%

±1.21
9/50
SRE Bench

2.29%

±0.93
24/35
Tax Agent Bench

32.62%

±3.38
19/80
Vibe Code Bench v1.1

77.48%

±4.04
36/123
BioMysteryBench

69.26%

±2.59
15/29
HLAB-Vals-1.1

8.75%

±2.19
22/33
IOI

59.33%

±4.60
18/46
LegalBench

83.76%

±0.41
50/151
ProgramBench

0.00%

±0.00
32/64
Terminal-Bench 4.0

8.08%

±1.34
46/59
Terminal-Bench-Science

1.43%

±1.43
38/43
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Tencent
Temperature: 0.9
Top P: 1
Top K: Default
Max Output Tokens: 64,000

Compared with

Hy4 Preview vs Kimi K3 21 shared benchmarks Hy4 Preview vs Step 5 Preview 17 shared benchmarks Hy4 Preview vs Muse Spark 1.2 14 shared benchmarks Hy4 Preview vs Mistral Large 4 14 shared benchmarks Hy4 Preview vs Ember-1 10 shared benchmarks Hy4 Preview vs GPT-6 Luna 21 shared benchmarks Hy4 Preview vs DeepSeek V4 Flash 0731 13 shared benchmarks Hy4 Preview vs Qwen 3.8 Max 14 shared benchmarks Hy4 Preview vs DeepSeek V4.1 Flash 19 shared benchmarks Hy4 Preview vs Gemini 3.7 Flash 13 shared benchmarks

Updates

Sep 17, 2026

We evaluated Tencent’s open-weight Hy4 Preview across our benchmark suite.

  • #21 of 58 on the Vals Index (55.40%) at $1.28 per test, the #4 open-weight model behind DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3.

  • Best open-weight model we’ve tested on Code Migration (47.43%, #7 of 61, at $3.41 per test versus $24.91 for GLM 5.3), Public Benefits Bench (68.61%, #4 of 36, behind only three Anthropic models) and SRE Bench (2.29%, #7 of 10).

  • Strong on reasoning and legal work: 75.00% on ProofBench v1.1 (#8 of 32), 59.33% on IOI (#10 of 28), 45.19% on Legal Research Bench (#10 of 61) and 9.17% on Harvey’s Legal Agent Benchmark (#13 of 62). Weaker on Terminal-Bench 2.1 (55.06%, #42 of 66), MedCode (43.25%, #37 of 93) and MedScribe (83.60%, #35 of 95).

  • Cost per test is below the board median on nearly every benchmark, skewing higher on benchmarks that involve files. Latency is medium-high for a model of its size and price (measured at low concurrency): closest to peers on math and science, slowest on routine knowledge work such as MedCode, Legal Research and Public Benefits Bench.

Hy4 Preview is a reasoning model with a 1M-token context window and 64k max output tokens, priced at $0.83 per million input tokens and $2.50 per million output tokens. We accessed it through OpenRouter.

Congrats to the Tencent team on the release!