Compare models
Pick up to 4 models for a side-by-side spec sheet. Hover any dotted metric name for what the test measures. Best value in each row is highlighted.
Claude Haiku 4.5DeepSeek V4 Pro
| Claude Haiku 4.5 Anthropic | DeepSeek V4 Pro DeepSeek | |
|---|---|---|
| Pricing | ||
| Input $/1M | $1.00 | $0.435 |
| Output $/1M | $5.00 | $0.87 |
| Blended $/1M (3:1) | $2.00 | $0.544 |
| Cache read $/1M | $0.10 | $0.0036 |
| Cache write $/1M | $1.25 | — |
| Quality indices | ||
| Intelligence Index | 29.6 | 44.3 |
| Coding Index | 43.9 | 59.4 |
| Math Index | — | — |
| Agentic Index | 16.4 | 36.4 |
| Speed & latency | ||
| Output speed (tok/s) | — | 74 |
| Time to first token | — | 0.99s |
| Time to first answer token | — | 60.41s |
| Benchmarks (Artificial Analysis) | ||
| GPQA Diamond | — | 88.8% |
| Humanity's Last Exam | — | 35.9% |
| SciCode | — | 50.0% |
| IFBench | — | 76.5% |
| AA-LCR (Long Context Reasoning) | — | 66.3% |
| Terminal-Bench Hard | — | 46.2% |
| Terminal-Bench 2.1 | — | 64.0% |
| τ²-Bench (Telecom) | — | 96.2% |
| τ³-Bench Banking | — | 25.8% |
| Design Arena (Elo) | ||
| Website | 1,149 | 1,261 |
| Web apps | — | 1,005 |
| Full-stack | — | 948 |
| UI components | 1,142 | 1,263 |
| Data viz | 1,160 | 1,212 |
| SVG | 1,072 | 1,183 |
| 3D | 1,131 | 1,310 |
| Game dev | 1,155 | 1,288 |
| Godot game dev | — | 1,059 |
| Code categories | 1,148 | 1,272 |
| ASCII art | 1,182 | 1,194 |
| Specs | ||
| Context window | 200K | 1.05M |
| Max output tokens | 64K | 384K |
| Input modalities | text, image, file | text |
| Tokenizer | Claude | DeepSeek |