Model Comparison
ERNIE 5.0 Thinking Preview
vs. Gemini 3.1 Flash-Lite
Comparing 2 AI models · 10 benchmarks · Baidu, Google
Recommended Pick
Go Gemini 3.1 Flash-Lite 8 metric wins Strongest on: Throughput, Intelligence, Coding
Lowest Price
ERNIE 5.0 Thinking Preview
$0.00/1M input price
Best Reasoning
ERNIE 5.0 Thinking Preview
57.9 reasoning score
Blends available reasoning benchmarks
Best for Coding
Gemini 3.1 Flash-Lite
30.1 coding index
Composite Indices
Higher is better; speed and price are normalized
Standard Benchmarks
Only benchmarks with data are shown
Differences That Matter
Price gap
ERNIE 5.0 Thinking Preview is ∞x cheaper on input tokens than Gemini 3.1 Flash-Lite.
Reasoning gap
ERNIE 5.0 Thinking Preview leads Gemini 3.1 Flash-Lite by 13.9 points on reasoning.
Coding gap
Gemini 3.1 Flash-Lite leads ERNIE 5.0 Thinking Preview by 0.9 points on coding.
Top-pick rationale
Gemini 3.1 Flash-Lite wins 8 measurable categories, including Throughput, Intelligence, Coding, GPQA.
Response Face-Off
Run one prompt through the selected models and compare response quality with live speed and cost context.
ERNIE 5.0 Thinking Preview
Baidu
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Gemini 3.1 Flash-Lite
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Which answer was more useful?
Full Comparison
| Metric | Ba ERNIE 5.0 Thinking Preview | Top Pick Go Gemini 3.1 Flash-Lite |
|---|---|---|
| Pricing per 1M tokens | ||
| Input Cost | $0.00/1M | $0.25/1M |
| Output Cost | $0.00/1M | $1.50/1M |
| Blended (3:1) | — | $0.56/1M |
| Specifications | ||
| Organization | Baidu | |
| Release Date | Nov 13, 2025 | Mar 3, 2026 |
| Performance & Speed | ||
| Throughput | — | 310.4 tok/s |
| TTFT | — | 5199ms |
| Latency | — | 5199ms |
| Composite Indices | ||
| Value Score | — | 100.0 |
| Reasoning Score | 57.9 | 44.0 |
| Intelligence | 29.1 | 33.5 |
| Coding | 29.2 | 30.1 |
| Math | 85.0 | — |
| Standard Benchmarks | ||
| GPQA | 77.7% | 82.2% |
| MMLU Pro | 83.0% | — |
| HLE | 12.7% | 16.2% |
| LiveCodeBench | 81.2% | — |
| AIME 2025 | 85.0% | — |
| SciCode | 37.5% | 41.9% |
| LCR | 6.7% | 65.3% |
| IFBench | 41.4% | 77.2% |
| TAU-bench v2 | 83.9% | 31.3% |
| TerminalBench Hard | 25.0% | 24.2% |
Key Takeaways
ERNIE 5.0 Thinking Preview offers the best value at $0.00/1M, making it ideal for high-volume applications and cost-conscious projects.
ERNIE 5.0 Thinking Preview has the strongest reasoning profile with a 57.9 reasoning score, combining the available reasoning-heavy benchmarks.
Gemini 3.1 Flash-Lite reaches a 30.1 coding index, making it the top choice for software development and code generation tasks.
All models support context windows of ∞+ tokens, suitable for processing lengthy documents and maintaining extended conversations.
When to Choose Each Model
ERNIE 5.0 Thinking Preview
- Cost-sensitive applications
- High-volume processing
- Complex reasoning tasks
- Research & analysis
Gemini 3.1 Flash-Lite
- Code generation
- Software development
