Compare · ModelsLive · 2 picked · head to head
DeepSeek V3.2 vs ERNIE 4.5 21B A3B Thinking
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
DeepSeek V3.2 wins on 6/6 benchmarks
DeepSeek V3.2 wins 6 of 6 shared benchmarks. Leads in math · knowledge · language.
Category leads
math·DeepSeek V3.2knowledge·DeepSeek V3.2language·DeepSeek V3.2coding·DeepSeek V3.2
Hype vs Reality
Attention vs performance
DeepSeek V3.2
#82 by perf·no signal
ERNIE 4.5 21B A3B Thinking
#55 by perf·no signal
Best value
ERNIE 4.5 21B A3B Thinking
2.0x better value than DeepSeek V3.2
DeepSeek V3.2
165.6 pts/$
$0.32/M
ERNIE 4.5 21B A3B Thinking
325.7 pts/$
$0.18/M
Vendor risk
Mixed exposure
One or more vendors flagged
DeepSeek
$3.4B·Tier 1
Baidu
$45.0B·Tier 1
Head to head
6 benchmarks · 2 models
DeepSeek V3.2ERNIE 4.5 21B A3B Thinking
OpenCompass · AIME2025
DeepSeek V3.2 leads by +16.8
DeepSeek V3.2
93.0
ERNIE 4.5 21B A3B Thinking
76.2
OpenCompass · GPQA-Diamond
DeepSeek V3.2 leads by +24.4
DeepSeek V3.2
84.6
ERNIE 4.5 21B A3B Thinking
60.2
OpenCompass · HLE
DeepSeek V3.2 leads by +16.7
DeepSeek V3.2
23.2
ERNIE 4.5 21B A3B Thinking
6.5
OpenCompass · IFEval
DeepSeek V3.2 leads by +8.5
DeepSeek V3.2
89.7
ERNIE 4.5 21B A3B Thinking
81.2
OpenCompass · LiveCodeBenchV6
DeepSeek V3.2 leads by +28.4
DeepSeek V3.2
75.4
ERNIE 4.5 21B A3B Thinking
47.0
OpenCompass · MMLU-Pro
DeepSeek V3.2 leads by +15.0
DeepSeek V3.2
85.8
ERNIE 4.5 21B A3B Thinking
70.8
Full benchmark table
| Benchmark | DeepSeek V3.2 | ERNIE 4.5 21B A3B Thinking |
|---|---|---|
OpenCompass · AIME2025 | 93.0 | 76.2 |
OpenCompass · GPQA-Diamond | 84.6 | 60.2 |
OpenCompass · HLE | 23.2 | 6.5 |
OpenCompass · IFEval | 89.7 | 81.2 |
OpenCompass · LiveCodeBenchV6 | 75.4 | 47.0 |
OpenCompass · MMLU-Pro | 85.8 | 70.8 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.26 | $0.38 | 164K tokens (~82 books) | $2.90 | |
| $0.07 | $0.28 | 131K tokens (~66 books) | $1.23 |