GPT-5 nano (high) vs GPT-5 nano (medium)
Comparing 2 AI models · 5 benchmarks · OpenAI
Composite Indices
Intelligence, Coding, Math
Standard Benchmarks
Academic and industry benchmarks
Benchmark Winners
GPT-5 nano (high)
- GPQA
- MMLU Pro
- HLE
- LiveCodeBench
- AIME 2025
GPT-5 nano (medium)
No clear wins
| Metric | Op GPT-5 nano (high) | Op GPT-5 nano (medium) |
|---|---|---|
| Pricing Per 1M tokens | ||
| Input Cost | $0.05/1M | $0.05/1M |
| Output Cost | $0.40/1M | $0.40/1M |
| Blended Cost 3:1 input/output ratio | $0.14/1M | $0.14/1M |
| Specifications | ||
| Organization Model creator | OpenAI | OpenAI |
| Release Date Launch date | Aug 7, 2025 | Aug 7, 2025 |
| Performance & Speed | ||
| Throughput Output speed | — | — |
| Time to First Token (TTFT) Initial response delay | — | — |
| Latency Time to first answer token | — | — |
| Composite Indices | ||
| Intelligence Index Overall reasoning capability | 51.0 | 49.3 |
| Coding Index Programming ability | 42.3 | 42.1 |
| Math Index Mathematical reasoning | 83.7 | 78.3 |
| Standard Benchmarks | ||
| GPQA Graduate-level reasoning | 67.6% | 67.0% |
| MMLU Pro Advanced knowledge | 78.0% | 77.2% |
| HLE Hard language evaluation | 8.2% | 7.6% |
| LiveCodeBench Real-world coding tasks | 78.9% | 76.3% |
| MATH 500 Mathematical problems | — | — |
| AIME 2025 Advanced math competition | 83.7% | 78.3% |
| AIME (Original) Math olympiad problems | — | — |
| SciCode Scientific code generation | 36.6% | 33.8% |
| LCR Code review capability | 41.7% | 40.0% |
| IFBench Instruction-following | 67.6% | 65.9% |
| TAU-bench v2 Tool use & agentic tasks | 36.5% | 30.4% |
| TerminalBench Hard CLI command generation | 11.3% | 16.3% |
Key Takeaways
GPT-5 nano (high) offers the best value at $0.05/1M, making it ideal for high-volume applications and cost-conscious projects.
GPT-5 nano (high) leads in reasoning capabilities with a 67.6% GPQA score, excelling at complex analytical tasks and problem-solving.
GPT-5 nano (high) achieves a 42.3 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
GPT-5 nano (high)
- Cost-sensitive applications
- High-volume processing
- Complex reasoning tasks
- Research & analysis
- Code generation
- Software development
GPT-5 nano (medium)
- General-purpose AI
- Versatile applications