The AI arena is free today

Open Superagent

DeepSeek-V3.2 (Thinking) vs Hy4 preview

Hy4 preview leads the LLM Stats Score 51.5 to 33.0.

DeepSeek · Tencent · Updated for 2026

Which is better?

Hy4 preview leads the overall LLM Stats Score 51.5 to 33.0, ranking #10 overall.

In the 3 individual benchmarks reported for both models, Hy4 preview wins 3; this is a narrower head-to-head signal than the composite indexes.

Based on current LLM Stats indexes, shared benchmarks, pricing, and model metadata for 2026.

Choose DeepSeek-V3.2 (Thinking)

  • you want predictable pricing at $0.28/M input and $0.42/M output

Choose Hy4 preview

  • overall performance matters — it scores 51.5 and ranks #10 on LLM Stats
  • your work emphasizes reasoning and coding — it leads those capability indexes
  • you value its reported benchmark strengths — it wins 3 of 3 exact shared results
  • you want the most recent training data — it shipped Aug 2026

At a glance

The differences that matter most.

Core performance indexes
33.0
#90
51.5
#10
33.0
#90
51.2
#8
22.9
#69
40.9
#10
11.1
#97
37.1
#11
Cost, coverage & limits
Benchmark wins
0 of 3
3 of 3
Input price
$0.28 / M
— / M
Output price
$0.42 / M
— / M
Context window
131,072

Capability indexes

Additional strengths measured across groups of related public benchmarks

2 shared
Index
DeepSeek-V3.2 (Thinking)
Hy4 preview
30.6#70
40.3#12
9.4#124
32.6#7
Conservative TrueSkill rating · higher is betterHow scores work

Individual benchmarks

14 reported for DeepSeek-V3.2 (Thinking) · 32 for Hy4 preview

3 shared

DeepSeek-V3.2 (Thinking) outperforms in 0 benchmarks, while Hy4 preview is better at 3 benchmarks (GPQA, SWE-bench Multilingual, Toolathlon).

Hy4 preview significantly outperforms across most benchmarks.

Tue Sep 01 2026 • llm-stats.com

Human preference

Blind head-to-head votes and playground preference scores

Model Size

Parameter count comparison

85.0B diff

Hy4 preview has 85.0B more parameters than DeepSeek-V3.2 (Thinking), making it 12.4% larger.

DeepSeek
DeepSeek-V3.2 (Thinking)
685.0Bparameters
Tencent
Hy4 preview
770.0Bparameters
685.0B
DeepSeek-V3.2 (Thinking)
770.0B
Hy4 preview

Context Window

Maximum input and output token capacity

Only DeepSeek-V3.2 (Thinking) specifies input context (131,072 tokens). Only DeepSeek-V3.2 (Thinking) specifies output context (65,536 tokens).

DeepSeek
DeepSeek-V3.2 (Thinking)
Input131,072 tokens
Output65,536 tokens
Tencent
Hy4 preview
Input- tokens
Output- tokens
Tue Sep 01 2026 • llm-stats.com

License

Usage and distribution terms

DeepSeek-V3.2 (Thinking) is licensed under MIT, while Hy4 preview uses Apache 2.0.

License differences may affect how you can use these models in commercial or open-source projects.

DeepSeek-V3.2 (Thinking)

MIT

Open weights

Hy4 preview

Apache 2.0

Open weights

Release Timeline

When each model was launched

DeepSeek-V3.2 (Thinking) was released on 2025-12-01, while Hy4 preview was released on 2026-08-28.

Hy4 preview is 9 months newer than DeepSeek-V3.2 (Thinking).

DeepSeek-V3.2 (Thinking)

Dec 1, 2025

9 months ago

Hy4 preview

Aug 28, 2026

3 days ago

9mo newer

Knowledge Cutoff

When training data ends

Neither model specifies a knowledge cutoff date.

Unable to compare the recency of their training data.

No cutoff dates available

Outputs Comparison

Notice missing or incorrect data?Start an Issue discussion

Judge for yourself.

Run your own prompts against DeepSeek-V3.2 (Thinking) and Hy4 preview side-by-side, then vote on the output you prefer.

DeepSeek-V3.2 (Thinking)
✓ Preferred
Hy4 preview
Open in Playground

FAQ

Common questions about DeepSeek-V3.2 (Thinking) vs Hy4 preview.

Which is better, DeepSeek-V3.2 (Thinking) or Hy4 preview?

Hy4 preview leads the LLM Stats Score 51.5 to 33.0. DeepSeek-V3.2 (Thinking) is made by DeepSeek and Hy4 preview is made by Tencent. The best choice depends on your use case — compare their capability indexes, individual benchmarks, pricing, and limits above.

How does DeepSeek-V3.2 (Thinking) compare to Hy4 preview in benchmarks?

DeepSeek-V3.2 (Thinking) scores AIME 2025: 93.1%, HMMT 2025: 90.2%, MMLU-Pro: 85.0%, LiveCodeBench: 83.3%, GPQA: 82.4%. Hy4 preview scores GPQA: 92.3%, Terminal-Bench 2.1: 85.4%, WideSearch: 83.9%, MCP Atlas: 83.7%, SWE-bench Multilingual: 82.9%.

What are the context window sizes for DeepSeek-V3.2 (Thinking) and Hy4 preview?

DeepSeek-V3.2 (Thinking) supports 131K tokens and Hy4 preview supports an unknown number of tokens. A larger context window lets you process longer documents, conversations, or codebases in a single request.

What are the main differences between DeepSeek-V3.2 (Thinking) and Hy4 preview?

Key differences include LLM Stats Score (33.0 vs 51.5), licensing (MIT vs Apache 2.0). See the full comparison above for benchmark-by-benchmark results.

Who makes DeepSeek-V3.2 (Thinking) and Hy4 preview?

DeepSeek-V3.2 (Thinking) is developed by DeepSeek and Hy4 preview is developed by Tencent.