Compare/DeepSeek V3.2 (Reasoning) vs Hy3-preview (Non-reasoning)

DeepSeek V3.2 (Reasoning)vsHy3-preview (Non-reasoning)

Side-by-side comparison of pricing, 12 benchmarks, and generation speed.

DeepSeek

DeepSeek V3.2 (Reasoning)

Input
$0.3/M
Output
$0.45/M
Speed
TTFT
Tencent

Hy3-preview (Non-reasoning)

Input
$0.123/M
Output
$0.43/M
Speed
120 tok/s
TTFT
2.86s

Winner by Category

Cheaper
Hy3-preview (Non-reasoning)
Faster (tok/s)
Hy3-preview (Non-reasoning)
Lower Latency
DeepSeek V3.2 (Reasoning)
Benchmarks (10-1)
DeepSeek V3.2 (Reasoning)

Pricing Comparison

MetricDeepSeek V3.2 (Reasoning)Hy3-preview (Non-reasoning)
Input ($/M tokens)$0.3$0.123
Output ($/M tokens)$0.45$0.43
Cost for 1M input + 100K output tokens:
DeepSeek V3.2 (Reasoning)$0.34
Hy3-preview (Non-reasoning)$0.17

Speed Comparison

Output Speed (tokens/s) — higher is better
DeepSeek V3.2 (Reasoning)
Hy3-preview (Non-reasoning)
120 tok/s
Time to First Token (seconds) — lower is better
DeepSeek V3.2 (Reasoning)
Hy3-preview (Non-reasoning)
2.86s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
41.733.7
Coding Index
36.734.3
Math Index
92.0
GPQA Diamond
84.0%73.2%
MMLU-Pro
86.2%
LiveCodeBench
86.2%
AIME 2025
92.0%
MATH-500
Humanity's Last Exam
22.2%6.3%
SciCode
38.9%39.4%
IFBench
60.7%48.0%
TerminalBench
35.6%31.8%
DeepSeek V3.2 (Reasoning)10 wins
1 winsHy3-preview (Non-reasoning)

Frequently Asked Questions

Which is cheaper, DeepSeek V3.2 (Reasoning) or Hy3-preview (Non-reasoning)?

Hy3-preview (Non-reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.20/M tokens vs $0.34/M for DeepSeek V3.2 (Reasoning).

Which model performs better on benchmarks?

DeepSeek V3.2 (Reasoning) wins 10 out of 12 benchmarks compared to 1 for Hy3-preview (Non-reasoning). See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

Hy3-preview (Non-reasoning) generates tokens faster at 120 tok/s vs 0 tok/s. DeepSeek V3.2 (Reasoning) also has lower time-to-first-token (0.00s vs 2.86s).

When should I use DeepSeek V3.2 (Reasoning) vs Hy3-preview (Non-reasoning)?

Choose based on your priorities: Hy3-preview (Non-reasoning) for lower cost, DeepSeek V3.2 (Reasoning) for stronger benchmark performance, and Hy3-preview (Non-reasoning) for faster generation. For latency-sensitive apps, check the TTFT comparison above.