Compare / head-to-head
Claude Opus 4.5vsClaude Opus 5.5
Claude Opus 5.5 leads 19 of 20 shared benchmarks. Claude Opus 5.5 is 1.3x cheaper per token. Claude Opus 5.5 has the larger context window (1M tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
1 – 19
Claude Opus 5.5 leads
Cheaper per token
Claude Opus 5.5
1.3x cheaper, input + output
Larger context
Claude Opus 5.5
1M tokens
Providers
Anthropic
same provider
Claude Opus 4.5
Anthropic · released 2025-11-24
textvision
Claude Opus 5.5
Anthropic · released 2026-09-22
textvision
Quality
Benchmark matrix
20 shared · 18 only Claude Opus 4.5 · 14 only Claude Opus 5.5| Benchmark | Claude Opus 4.5 | Claude Opus 5.5 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 20 | ||||
| GPQA Diamond | 87% ↗ | 90.6% ↗epoch run | -3.6 pt | Claude Opus 5.5 |
| LMArena Elo | 1473.6 ↗ | 1503.7 ↗ | -30.1 | Claude Opus 5.5 |
| EBR-bench (Epoch AI run) | 14.3% ↗epoch run | 71.4% ↗epoch run | -57.1 pt | Claude Opus 5.5 |
| FrontierMath Tier 4 v2 (Epoch AI run) | 4.9% ↗epoch run | 95% ↗epoch run | -90.1 pt | Claude Opus 5.5 |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 34.4% ↗epoch run | 91.2% ↗epoch run | -56.8 pt | Claude Opus 5.5 |
| Furniture Assembly (Epoch AI run) | 28.3% ↗epoch run | 83.3% ↗epoch run | -55 pt | Claude Opus 5.5 |
| LiveBench Agentic Coding (LiveBench) | 39.7% ↗ | 71.7% ↗ | -32 pt | Claude Opus 5.5 |
| LiveBench Coding (LiveBench) | 79.7% ↗ | 89.3% ↗ | -9.6 pt | Claude Opus 5.5 |
| LiveBench Data Analysis (LiveBench) | 74.4% ↗ | 80.3% ↗ | -5.9 pt | Claude Opus 5.5 |
| LiveBench Instruction Following (LiveBench) | 62.5% ↗ | 67% ↗ | -4.5 pt | Claude Opus 5.5 |
| LiveBench Language (LiveBench) | 81.3% ↗ | 86.3% ↗ | -5 pt | Claude Opus 5.5 |
| LiveBench Mathematics (LiveBench) | 90.4% ↗ | 97.1% ↗ | -6.7 pt | Claude Opus 5.5 |
| LiveBench Reasoning (LiveBench) | 80.1% ↗ | 92.2% ↗ | -12.1 pt | Claude Opus 5.5 |
| LMArena WebDev (LMArena) | 1493.3 ↗ | 1815.4 ↗ | -322.1 | Claude Opus 5.5 |
| Mystery Game Puzzles (Epoch AI run) | 22% ↗epoch run | 71% ↗epoch run | -49 pt | Claude Opus 5.5 |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 86.1% ↗epoch run | 100% ↗epoch run | -13.9 pt | Claude Opus 5.5 |
| SAGE (Vals AI) | 52.1% ↗ | 45.8% ↗ | +6.3 pt | Claude Opus 4.5 |
| SimpleBench (SimpleBench) | 62% ↗ | 88.4% ↗ | -26.4 pt | Claude Opus 5.5 |
| SimpleQA Verified | 45.7% ↗epoch run | 72.2% ↗epoch run | -26.5 pt | Claude Opus 5.5 |
| Vending-Bench 2 (Andon Labs) | 4967.06 ↗ | 9235.25 ↗ | -4268.2 | Claude Opus 5.5 |
| Only Claude Opus 4.5 reports · 18 | ||||
| SWE-bench Verified | 80.9% ↗ | not reported | — | — |
| ARC-AGI-2 (Verified) | 37.6% ↗ | not reported | — | — |
| BALROG (BALROG) | 43.5% ↗ | not reported | — | — |
| Chess Puzzles (Epoch AI run) | 12% ↗epoch run | not reported | — | — |
| GSO Opt@1 (GSO) | 24.5% ↗ | not reported | — | — |
| MCP Atlas | 62.3% ↗ | not reported | — | — |
| MMMLU | 90.8% ↗ | not reported | — | — |
| MMMU (validation) | 80.7% ↗ | not reported | — | — |
| OSWorld | 66.3% ↗ | not reported | — | — |
| SWE-bench Verified (Epoch AI run) | 76.7% ↗epoch run | not reported | — | — |
| tau2-bench Airline (Sierra) | 84% ↗ | not reported | — | — |
| tau2-bench Banking Knowledge (Sierra) | 24.7% ↗ | not reported | — | — |
| tau2-bench Retail (Sierra) | 79.6% ↗ | not reported | — | — |
| tau2-bench Telecom (Sierra) | 92.3% ↗ | not reported | — | — |
| Terminal-Bench 2.0 | 59.3% ↗ | not reported | — | — |
| WeirdML (Håvard Tveit Ihle) | 63.7% ↗ | not reported | — | — |
| τ2-bench (Retail) | 88.9% ↗ | not reported | — | — |
| τ2-bench (Telecom) | 98.2% ↗ | not reported | — | — |
| Only Claude Opus 5.5 reports · 14 | ||||
| APEX-Agents (Mercor) | not reported | 73.5% ↗ | — | — |
| AutomationBench 1.0.6 | not reported | 40% ↗ | — | — |
| Chartography (with tools) | not reported | 89% ↗ | — | — |
| CursorBench 4.0 | not reported | 57.8% ↗ | — | — |
| FrontierCode v1.1 (Main) | not reported | 54.4% ↗ | — | — |
| FrontierSWE V2 (Proximal Labs) | not reported | 62.3% ↗ | — | — |
| GDPval-AA v2.1 | not reported | 1846 ↗ | — | — |
| Humanity's Last Exam (with tools) | not reported | 67.7% ↗ | — | — |
| LMArena Agent (LMArena) | not reported | 0.1382 ↗ | — | — |
| MirrorCode (Epoch AI run) | not reported | 77.4% ↗epoch run | — | — |
| OSWorld 2.0 (partial) | not reported | 81.8% ↗ | — | — |
| Terminal-Bench 4.0 | not reported | 66.4% ↗ | — | — |
| Terminal-Bench 4.0 (Vals AI) | not reported | 65.2% ↗ | — | — |
| Terminal-Bench Science 0.1 | not reported | 58.7% ↗ | — | — |
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 4.5 minus Claude Opus 5.5 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 4.5 | Claude Opus 5.5 | Edge |
|---|---|---|---|
| Context window | 200K tokens | 1M tokens | Claude Opus 5.5 |
| Max output | 64K tokens | 128K tokens | Claude Opus 5.5 |
| Input price / 1M | $5 ↗ | $4 ↗ | Claude Opus 5.5 |
| Output price / 1M | $25 ↗ | $20 ↗ | Claude Opus 5.5 |
| Cached input / 1M | $0.5 ↗ | $0.2 ↗ | Claude Opus 5.5 |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $30 | $24 | Claude Opus 5.5 |
| Modalities | text · vision | text · vision | Tie |
| Released | 2025-11-24 | 2026-09-22 | — |
| Cited benchmark scores | 39 | 34 | — |
Reliability
Anthropic status
Live from /statusMore matchups
Claude Opus 4.5 vs …
Models sharing the most benchmarksMore matchups
Claude Opus 5.5 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Claude Opus 4.5 and Claude Opus 5.5 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.