Compare models. Find the differences.
Add, remove, or change models in the table.
Benchmark snapshot
2 benchmarks in common.
4 models selected. 2 benchmarks in common.
Reported scores; test settings may differ.
| Benchmark | OpenAI | Anthropic | Qwen | MiniMax |
|---|---|---|---|---|
| Arena ELOHuman preference | 1,437 | 1,426 | 1,394 | 1,391 |
| GPQA DiamondScientific knowledge | Not in our dataset | 79.6% | 84.2% | 85.2% |
| HLEAcademic reasoning | 27.8% | 10.7% | 22.4% | 19.4% |
| MMLU ProMulti-task understanding | Not in our dataset | 85.0% | 85.3% | Not in our dataset |
| AIME 2025Mathematics | Not in our dataset | 75.5% | Not in our dataset | 86.3% |
| GSOGraduate science | Not in our dataset | 6.9% | Not in our dataset | Not in our dataset |
| AiderCode editing | Not in our dataset | 72.0% | Not in our dataset | Not in our dataset |
| SWE-Bench VerifiedAgentic coding | Not in our dataset | 72.5% | 69.2% | Not in our dataset |
| LiveCodeBenchCompetitive coding | Not in our dataset | 56.6% | Not in our dataset | Not in our dataset |
| Tau-BenchAgentic tool use | Not in our dataset | 81.4% | Not in our dataset | Not in our dataset |
| Terminal-BenchAgentic terminal | 54.0% | Not in our dataset | Not in our dataset | 42.2% |
| ARC-AGI-2Visual reasoning | 0.8% | 8.6% | Not in our dataset | 4.9% |
Pricing & context Published USD rates per 1M tokens
| Specification | GPT-5.2 | Claude Opus 4 4.0 Thinking | 3.5 35B A3B Thinking | M2 M2.5 |
|---|---|---|---|---|
| Input | $1.75 | $15 | $0.14 | $0.255 |
| Output | $14 | $75 | $1 | $1.02 |
| Context window | 400K | 200K | 262K | 205K |
| Pricing details | Explore pricing → | Explore pricing → | Explore pricing → | Explore pricing → |
How to read this comparison
Each column keeps the exact model version you selected. Shared benchmarks are the intersection of their reported results; “all coverage” also shows where evidence is missing. Missing results never count as zero.
These are published benchmark scores, not a controlled head-to-head test. The available evidence does not establish matching evaluation settings, so we show no cross-model winner highlights or deltas. Composite quality scores are omitted because they can summarize different benchmark sets.
Scores and statistics are computed upstream. This page reads a fixed data snapshot; its date describes the snapshot, not the date of every measurement.