Compare / head-to-head
Claude Opus 4.7vs
GPT-5.2-Codex
Claude Opus 4.7 leads 9 of 10 shared benchmarks. GPT-5.2-Codex is 1.9x cheaper per token. Claude Opus 4.7 has the larger context window (1M tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
9 – 1
Claude Opus 4.7 leads
Cheaper per token
GPT-5.2-Codex
1.9x cheaper, input + output
Larger context
Claude Opus 4.7
1M tokens
Providers
2 providers
Anthropic · OpenAI
Claude Opus 4.7
Anthropic · released 2026-04-16
textvision
GPT-5.2-Codex
OpenAI · released 2025-12-18
textvision
Quality
Benchmark matrix
10 shared · 28 only Claude Opus 4.7 · 3 only GPT-5.2-Codex| Benchmark | Claude Opus 4.7 | GPT-5.2-Codex | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 10 | ||||
| LiveBench Agentic Coding (LiveBench) | 50.7% ↗ | 49.4% ↗ | +1.3 pt | Claude Opus 4.7 |
| LiveBench Coding (LiveBench) | 82.1% ↗ | 83.6% ↗ | -1.5 pt | GPT-5.2-Codex |
| LiveBench Data Analysis (LiveBench) | 78.3% ↗ | 78.2% ↗ | +0.1 pt | Claude Opus 4.7 |
| LiveBench Instruction Following (LiveBench) | 66.7% ↗ | 66.4% ↗ | +0.3 pt | Claude Opus 4.7 |
| LiveBench Language (LiveBench) | 77.9% ↗ | 73.7% ↗ | +4.2 pt | Claude Opus 4.7 |
| LiveBench Mathematics (LiveBench) | 92.9% ↗ | 88.8% ↗ | +4.1 pt | Claude Opus 4.7 |
| LiveBench Reasoning (LiveBench) | 87.2% ↗ | 77.7% ↗ | +9.5 pt | Claude Opus 4.7 |
| LMArena WebDev (LMArena) | 1557.3 ↗ | 1338.9 ↗ | +218.4 | Claude Opus 4.7 |
| OSWorld-Verified | 82.8% ↗ | 38.2% ↗ | +44.6 pt | Claude Opus 4.7 |
| SWE-Bench Pro | 64.3% ↗ | 56.4% ↗ | +7.9 pt | Claude Opus 4.7 |
| Only Claude Opus 4.7 reports · 28 | ||||
| SWE-bench Verified | 87.6% ↗ | not reported | — | — |
| GPQA Diamond | 94.2% ↗ | not reported | — | — |
| AIME 2026 | 95.8% ↗matharena ⚠ | not reported | — | — |
| Humanity's Last Exam (no tools) | 46.9% ↗ | not reported | — | — |
| LMArena Elo | 1483.4 ↗ | not reported | — | — |
| APEX-Agents (Mercor) | 49.2% ↗ | not reported | — | — |
| BrowseComp | 79.8% ↗ | not reported | — | — |
| Chess Puzzles (Epoch AI run) | 30% ↗epoch run | not reported | — | — |
| EBR-bench (Epoch AI run) | 19% ↗epoch run | not reported | — | — |
| FrontierMath Tier 4 v2 (Epoch AI run) | 31.7% ↗epoch run | not reported | — | — |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 70.2% ↗epoch run | not reported | — | — |
| Furniture Assembly (Epoch AI run) | 33.3% ↗epoch run | not reported | — | — |
| GSO Opt@1 (GSO) | 42.2% ↗ | not reported | — | — |
| Humanity's Last Exam (with tools) | 54.7% ↗ | not reported | — | — |
| LMArena Vision (LMArena) | 1298.1 ↗ | not reported | — | — |
| MirrorCode (Epoch AI run) | 31.1% ↗epoch run | not reported | — | — |
| Mystery Game Puzzles (Epoch AI run) | 28% ↗epoch run | not reported | — | — |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 97.8% ↗epoch run | not reported | — | — |
| SAGE (Vals AI) | 56.1% ↗ | not reported | — | — |
| SimpleBench (SimpleBench) | 61.7% ↗ | not reported | — | — |
| SimpleQA Verified | 51.7% ↗epoch run | not reported | — | — |
| SWE-bench Multilingual | 80.5% ↗ | not reported | — | — |
| SWE-bench Multimodal | 34.5% ↗ | not reported | — | — |
| SWE-bench Verified (Epoch AI run) | 83.5% ↗epoch run | not reported | — | — |
| tau2-bench Banking Knowledge (Sierra) | 40.2% ↗ | not reported | — | — |
| Terminal-Bench 2.1 | 66.1% ↗ | not reported | — | — |
| Vending-Bench 2 (Andon Labs) | 10936.76 ↗ | not reported | — | — |
| WeirdML (Håvard Tveit Ihle) | 76.4% ↗ | not reported | — | — |
| Only GPT-5.2-Codex reports · 3 | ||||
| Cybersecurity Capture The Flag Challenges | not reported | 67.4% ↗ | — | — |
| SWE-Lancer IC Diamond | not reported | 76% ↗ | — | — |
| Terminal-Bench 2.0 | not reported | 64% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Claude Opus 4.7 minus GPT-5.2-Codex in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Claude Opus 4.7 | GPT-5.2-Codex | Edge |
|---|---|---|---|
| Context window | 1M tokens | 400K tokens | Claude Opus 4.7 |
| Max output | 128K tokens | 128K tokens | Tie |
| Input price / 1M | $5 ↗ | $1.75 ↗ | GPT-5.2-Codex |
| Output price / 1M | $25 ↗ | $14 ↗ | GPT-5.2-Codex |
| Cached input / 1M | $0.5 ↗ | $0.175 ↗ | GPT-5.2-Codex |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $30 | $15.75 | GPT-5.2-Codex |
| Modalities | text · vision | text · vision | Tie |
| Released | 2026-04-16 | 2025-12-18 | — |
| Cited benchmark scores | 46 | 13 | — |
Reliability
Provider status
Live from /statusMore matchups
Claude Opus 4.7 vs …
Models sharing the most benchmarksMore matchups
GPT-5.2-Codex vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Claude Opus 4.7 and GPT-5.2-Codex on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.