Up to four models
Claude Opus 4 vs DeepSeek-R1-0528
On the index (GPT-4 = 100), Claude Opus 4 scores 136 and DeepSeek-R1-0528 135. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Claude Opus 4 and DeepSeek-R1-0528 are too close to call (0 points apart, within the uncertainty), ahead on 1 of 2 shared core tests.
Test by testthird-party results
Claude Opus 4
DeepSeek-R1-0528
Every result as a table
| Test | Claude Opus 4 | DeepSeek-R1-0528 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 76.3% | 76.3% |
| Fiction.LiveBench | 61.1% | 75.0% |
| SimpleBench | 58.8% | 40.8% |
| Maths | ||
| AIME-style maths (OTIS mock) | 64.4% | 66.4% |
| MATH (level 5) | 85.0% | 96.6% |
| Coding | ||
| Aider Polyglot | 72.0% | 71.4% |
| WeirdML | 43.7% | 41.6% |
| Agents | ||
| DeepResearch Bench | 46.8% | 35.1% |
Task length · METR38 sec
32 sec
Tests at least two of these models have taken. Bold rows are the core tests.