Gemini 2.5 Flash (Non-reasoning) vs Gemini 2.5 Flash Preview (Sep '25) (Reasoning)
Comparing 2 AI models · 6 benchmarks · Google
Composite Indices
Intelligence, Coding, Math
Standard Benchmarks
Academic and industry benchmarks
Benchmark Winners
Gemini 2.5 Flash (Non-reasoning)
- MATH 500
Gemini 2.5 Flash Preview (Sep '25) (Reasoning)
- GPQA
- MMLU Pro
- HLE
- LiveCodeBench
- AIME 2025
| Metric | Go Gemini 2.5 Flash (Non-reasoning) | Go Gemini 2.5 Flash Preview (Sep '25) (Reasoning) |
|---|---|---|
| Pricing Per 1M tokens | ||
| Input Cost | $0.30/1M | $0.30/1M |
| Output Cost | $2.50/1M | $2.50/1M |
| Blended Cost 3:1 input/output ratio | $0.85/1M | $0.85/1M |
| Specifications | ||
| Organization Model creator | ||
| Release Date Launch date | May 20, 2025 | Sep 25, 2025 |
| Performance & Speed | ||
| Throughput Output speed | 237.8 tok/s | 155.2 tok/s |
| Time to First Token (TTFT) Initial response delay | 401ms | 8247ms |
| Latency Time to first answer token | 401ms | 8247ms |
| Composite Indices | ||
| Intelligence Index Overall reasoning capability | 40.4 | 54.4 |
| Coding Index Programming ability | 30.0 | 42.5 |
| Math Index Mathematical reasoning | 60.3 | 78.3 |
| Standard Benchmarks | ||
| GPQA Graduate-level reasoning | 68.3% | 79.3% |
| MMLU Pro Advanced knowledge | 80.9% | 84.2% |
| HLE Hard language evaluation | 5.1% | 12.7% |
| LiveCodeBench Real-world coding tasks | 49.5% | 71.3% |
| MATH 500 Mathematical problems | 93.2% | — |
| AIME 2025 Advanced math competition | 60.3% | 78.3% |
| AIME (Original) Math olympiad problems | 50.0% | — |
| SciCode Scientific code generation | 29.1% | 40.5% |
| LCR Code review capability | 45.9% | 64.3% |
| IFBench Instruction-following | 39.0% | 52.3% |
| TAU-bench v2 Tool use & agentic tasks | 14.9% | 45.6% |
| TerminalBench Hard CLI command generation | 11.3% | 15.6% |
Key Takeaways
Gemini 2.5 Flash (Non-reasoning) offers the best value at $0.30/1M, making it ideal for high-volume applications and cost-conscious projects.
Gemini 2.5 Flash Preview (Sep '25) (Reasoning) leads in reasoning capabilities with a 79.3% GPQA score, excelling at complex analytical tasks and problem-solving.
Gemini 2.5 Flash Preview (Sep '25) (Reasoning) achieves a 42.5 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
Gemini 2.5 Flash (Non-reasoning)
- Cost-sensitive applications
- High-volume processing
Gemini 2.5 Flash Preview (Sep '25) (Reasoning)
- Complex reasoning tasks
- Research & analysis
- Code generation
- Software development
Try Models for Free
Try Gemini 2.5 Flash (Non-reasoning) for FREE
No credit card or account required.
Try Gemini 2.5 Flash Preview (Sep '25) (Reasoning) for FREE
No credit card or account required.
Cost Calculator
Gemini 2.5 Flash (Non-reasoning) Gemini 2.5 Flash Preview (Sep '25) (Reasoning) Costs are estimates based on API pricing. Actual costs may vary based on caching, batch processing, and volume discounts.