Price Per TokenPrice Per Token
Baidu
Baidu
vs
Google
Google

ERNIE 4.5 21B A3B Thinking vs Gemini 3 Pro Preview

A detailed comparison of pricing, benchmarks, and capabilities

Best Coding Model Leaderboard Find the top LLM for coding by benchmarks and community votes

114 out of our 298 tracked models have had a price change in February.

Get our weekly newsletter on pricing changes, new releases, and tools.

Key Takeaways

ERNIE 4.5 21B A3B Thinking wins:

  • Cheaper input tokens
  • Cheaper output tokens

Gemini 3 Pro Preview wins:

  • Larger context window
  • Faster response time
  • Higher intelligence benchmark
  • Better at coding
  • Better at math
  • Supports vision
Price Advantage
ERNIE 4.5 21B A3B Thinking
Benchmark Advantage
Gemini 3 Pro Preview
Context Window
Gemini 3 Pro Preview
Speed
Gemini 3 Pro Preview

Pricing Comparison

Price Comparison

MetricERNIE 4.5 21B A3B ThinkingGemini 3 Pro PreviewWinner
Input (per 1M tokens)$0.07$2.00 ERNIE 4.5 21B A3B Thinking
Output (per 1M tokens)$0.28$12.00 ERNIE 4.5 21B A3B Thinking
Cache Read (per 1M)N/A$200000.00 Gemini 3 Pro Preview
Cache Write (per 1M)N/A$375000.00 Gemini 3 Pro Preview
Using a 3:1 input/output ratio, ERNIE 4.5 21B A3B Thinking is 97% cheaper overall.

ERNIE 4.5 21B A3B Thinking Providers

Novita $0.07 (Cheapest)

Gemini 3 Pro Preview Providers

Google AI Studio $2.00 (Cheapest)
Google $2.00 (Cheapest)

Benchmark Comparison

6
Benchmarks Compared
0
ERNIE 4.5 21B A3B Thinking Wins
0
Gemini 3 Pro Preview Wins

Benchmark Scores

BenchmarkERNIE 4.5 21B A3B ThinkingGemini 3 Pro PreviewWinner
Intelligence Index
Overall intelligence score
-48.4-
Coding Index
Code generation & understanding
-46.5-
Math Index
Mathematical reasoning
-95.7-
MMLU Pro
Academic knowledge
-89.8-
GPQA
Graduate-level science
-90.8-
LiveCodeBench
Competitive programming
-91.7-
Gemini 3 Pro Preview significantly outperforms in coding benchmarks.

Cost vs Quality

X-axis:
Y-axis:
Loading chart...
Other models

Context & Performance

Context Window

ERNIE 4.5 21B A3B Thinking
131,072
tokens
Max output: 65,536 tokens
Gemini 3 Pro Preview
1,048,576
tokens
Max output: 65,536 tokens
Gemini 3 Pro Preview has a 88% larger context window.

Speed Performance

MetricERNIE 4.5 21B A3B ThinkingGemini 3 Pro PreviewWinner
Tokens/secondN/A139.9 tok/s
Time to First TokenN/A27.20s

Capabilities

Feature Comparison

FeatureERNIE 4.5 21B A3B ThinkingGemini 3 Pro Preview
Vision (Image Input)
Tool/Function Calls
Reasoning Mode
Audio Input
Audio Output
PDF Input
Prompt Caching
Web Search

License & Release

PropertyERNIE 4.5 21B A3B ThinkingGemini 3 Pro Preview
LicenseOpen SourceProprietary
AuthorBaiduGoogle
ReleasedOct 2025Nov 2025

ERNIE 4.5 21B A3B Thinking Modalities

Input
text
Output
text

Gemini 3 Pro Preview Modalities

Input
textimagefileaudiovideo
Output
text

Related Comparisons

Compare ERNIE 4.5 21B A3B Thinking with:

Compare Gemini 3 Pro Preview with:

Frequently Asked Questions

ERNIE 4.5 21B A3B Thinking has cheaper input pricing at $0.07/M tokens. ERNIE 4.5 21B A3B Thinking has cheaper output pricing at $0.28/M tokens.
Gemini 3 Pro Preview scores higher on coding benchmarks with a score of 46.5, compared to ERNIE 4.5 21B A3B Thinking's score of N/A.
ERNIE 4.5 21B A3B Thinking has a 131,072 token context window, while Gemini 3 Pro Preview has a 1,048,576 token context window.
ERNIE 4.5 21B A3B Thinking does not support vision. Gemini 3 Pro Preview supports vision.