GPT-5.2vsOpus 4.5
Across 13 shared benchmarks, GPT-5.2 leads overall: GPT-5.2 wins 9, Opus 4.5 wins 4, with 0 ties and an average score difference of +2.82.
GPT-5.29 wins(69%)(31%)4 winsOpus 4.5
Benchmark scores
Grouped by capability, sorted by largest gap within each. 13 shared benchmarks.
General Knowledge
GPT-5.2 4/5| Benchmark | GPT-5.2 | Opus 4.5 | Diff |
|---|---|---|---|
| LiveBench | 48.9194 / 115Normal (No Tools) | 75.9611 / 115Thinking (No Tools, 64K Budget) | -27.05 |
| ARC-AGI-2 | 54.2023 / 62深度思考（无工具、并行） | 37.6029 / 62Extended (no tools) | +16.60 |
| ARC-AGI | 90.5017 / 68深度思考（无工具、并行） | 8024 / 68Extended (no tools) | +10.50 |
| GPQA Diamond | 93.209 / 187深度思考（无工具、并行） | 8742 / 187Extended (no tools) | +6.20 |
| HLE | 45.5041 / 172Deep Thinking (With Tools + Internet) | 43.2049 / 172Extended (with tools) | +2.30 |
Agent Level Benchmark
GPT-5.2 2/2| Benchmark | GPT-5.2 | Opus 4.5 | Diff |
|---|---|---|---|
| τ²-Bench - Telecom | 98.704 / 35极高强度思考（工具） | 90.7021 / 35Extended (with tools) | +8 |
| τ²-Bench | 8212 / 43极高强度思考（工具） | 81.9913 / 43Extended (with tools) | +0.01 |
Math and Reasoning
GPT-5.2 2/2| Benchmark | GPT-5.2 | Opus 4.5 | Diff |
|---|---|---|---|
| FrontierMath | 40.308 / 60极高强度思考（工具） | 20.7017 / 60Extended (no tools) | +19.60 |
| FrontierMath - Tier 4 | 18.8016 / 80Thinking High (No Tools) | 4.2040 / 80Normal (No Tools) | +14.60 |
AI Agent - Tool Usage
Opus 4.5 1/1| Benchmark | GPT-5.2 | Opus 4.5 | Diff |
|---|---|---|---|
| MCP-Atlas | 67.6018 / 27极高强度思考（工具） | 69.8016 / 27Thinking High (With Tools) | -2.20 |
Coding and Software Engineer
Opus 4.5 1/1| Benchmark | GPT-5.2 | Opus 4.5 | Diff |
|---|---|---|---|
| SWE-bench Verified | 8017 / 112极高强度思考（工具） | 80.909 / 112Extended (with tools) | -0.90 |
Commonsense Reasoning
Opus 4.5 1/1| Benchmark | GPT-5.2 | Opus 4.5 | Diff |
|---|---|---|---|
| Simple Bench | 45.8033 / 63Thinking High (No Tools) | 6212 / 63Extended (no tools) | -16.20 |
Multimodal Understanding
GPT-5.2 1/1| Benchmark | GPT-5.2 | Opus 4.5 | Diff |
|---|---|---|---|
| MMMU | 85.901 / 29极高强度思考（无工具） | 80.7011 / 29Extended (no tools) | +5.20 |
Specs
| Field | GPT-5.2 | Opus 4.5 |
|---|---|---|
| Publisher | OpenAI | Anthropic |
| Release date | 2025-12-11 | 2025-11-25 |
| Model type | Chat model | Reasoning model |
| Architecture | Dense | Dense |
| Parameters | Not available | Not available |
| Context length | 400K | 200K |
| Max output | Not available | 64K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | GPT-5.2 | Opus 4.5 |
|---|---|---|
| Text input | $1.75 / 1M tokens | $5 / 1M tokens |
| Text output | $14 / 1M tokens | $25 / 1M tokens |
| Cache read | $0.175 / 1M tokens | $0.5 / 1M tokens |
| Cache write | $1.75 / 1M tokens | $6.25 / 1M tokens |
Summary
- GPT-5.2leads in:General Knowledge (4/5), Agent Level Benchmark (2/2), Math and Reasoning (2/2), Multimodal Understanding (1/1)
- Opus 4.5leads in:AI Agent - Tool Usage (1/1), Coding and Software Engineer (1/1), Commonsense Reasoning (1/1)
On average across the 13 shared benchmarks, GPT-5.2 scores 2.82 higher.
Largest single-benchmark gap: LiveBench — GPT-5.2 48.91 vs Opus 4.5 75.96 (-27.05).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.