Compare / head-to-head
Claude Opus 5vs
Grok 4
Claude Opus 5 leads 10 of 10 shared benchmarks. Claude Opus 5 has the larger context window (1M tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
10 – 0
Claude Opus 5 leads
Cheaper per token
—
list price, input + output
Larger context
Claude Opus 5
1M tokens
Providers
2 providers
Anthropic · xAI
Claude Opus 5
Anthropic · released 2026-07-24
textvision
Grok 4
xAI · released 2025-07-09
textvision
- Context
- 256K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 17 · 8 core
Quality
Benchmark matrix
10 shared · 26 only Claude Opus 5 · 7 only Grok 4| Benchmark | Claude Opus 5 | Grok 4 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 10 | ||||
| GPQA Diamond | 93.9% ↗epoch run | 87% ↗epoch run | +6.9 pt | Claude Opus 5 |
| Humanity's Last Exam (no tools) | 56.6% ↗ | 25.4% ↗ | +31.2 pt | Claude Opus 5 |
| LMArena Elo | 1489.7 ↗ | 1410.8 ↗ | +78.9 | Claude Opus 5 |
| BALROG (BALROG) | 63.4% ↗ | 43.6% ↗ | +19.8 pt | Claude Opus 5 |
| Chess Puzzles (Epoch AI run) | 42% ↗epoch run | 28% ↗epoch run | +14 pt | Claude Opus 5 |
| LMArena Vision (LMArena) | 1289 ↗ | 1184.2 ↗ | +104.8 | Claude Opus 5 |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 98.9% ↗epoch run | 84% ↗epoch run | +14.9 pt | Claude Opus 5 |
| SAGE (Vals AI) | 49.4% ↗ | 25.1% ↗ | +24.3 pt | Claude Opus 5 |
| SimpleBench (SimpleBench) | 80.6% ↗ | 60.5% ↗ | +20.1 pt | Claude Opus 5 |
| WeirdML (Håvard Tveit Ihle) | 91.8% ↗ | 45.7% ↗ | +46.1 pt | Claude Opus 5 |
| Only Claude Opus 5 reports · 26 | ||||
| APEX-Agents (Mercor) | 65.8% ↗ | not reported | — | — |
| CursorBench 3.2.0 | 70% ↗ | not reported | — | — |
| DeepSWE v1.1 (Datacurve) | 73.6% ↗ | not reported | — | — |
| EBR-bench (Epoch AI run) | 45.7% ↗epoch run | not reported | — | — |
| FrontierMath Tier 4 v2 (Epoch AI run) | 73.2% ↗epoch run | not reported | — | — |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 85.6% ↗epoch run | not reported | — | — |
| FrontierSWE V2 (Proximal Labs) | 52% ↗ | not reported | — | — |
| Furniture Assembly (Epoch AI run) | 60.8% ↗epoch run | not reported | — | — |
| GDPval-AA v2 | 1824 ↗ | not reported | — | — |
| LiveBench Agentic Coding (LiveBench) | 65.2% ↗ | not reported | — | — |
| LiveBench Coding (LiveBench) | 81.4% ↗ | not reported | — | — |
| LiveBench Data Analysis (LiveBench) | 74.6% ↗ | not reported | — | — |
| LiveBench Instruction Following (LiveBench) | 63.8% ↗ | not reported | — | — |
| LiveBench Language (LiveBench) | 88.7% ↗ | not reported | — | — |
| LiveBench Mathematics (LiveBench) | 95.7% ↗ | not reported | — | — |
| LiveBench Reasoning (LiveBench) | 91.2% ↗ | not reported | — | — |
| LMArena Agent (LMArena) | 0.0867 ↗ | not reported | — | — |
| LMArena WebDev (LMArena) | 1695 ↗ | not reported | — | — |
| Mystery Game Puzzles (Epoch AI run) | 59% ↗epoch run | not reported | — | — |
| OSWorld-Verified (XLANG) | 83.4% ↗ | not reported | — | — |
| SimpleQA Verified | 59.9% ↗epoch run | not reported | — | — |
| SWE-rebench 2026-05-15 to 2026-07-01 (Nebius) | 63.4% ↗ | not reported | — | — |
| tau2-bench Banking Knowledge (Sierra) | 48.7% ↗ | not reported | — | — |
| Terminal-Bench 4.0 (Vals AI) | 53.5% ↗ | not reported | — | — |
| Terminal-Bench Science 0.1 | 29% ↗ | not reported | — | — |
| Vending-Bench 2 (Andon Labs) | 11181.87 ↗ | not reported | — | — |
| Only Grok 4 reports · 7 | ||||
| AIME 2025 | not reported | 91.7% ↗ | — | — |
| ARC-AGI-2 | not reported | 15.9% ↗ | — | — |
| GPQA (no tools) | not reported | 87.5% ↗ | — | — |
| HMMT 2025 (no tools) | not reported | 90% ↗ | — | — |
| Humanity's Last Exam (with tools) | not reported | 38.6% ↗ | — | — |
| LiveCodeBench (Jan - May) | not reported | 79% ↗ | — | — |
| USAMO 2025 (no tools) | not reported | 37.5% ↗ | — | — |
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 5 minus Grok 4 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 5 | Grok 4 | Edge |
|---|---|---|---|
| Context window | 1M tokens | 256K tokens | Claude Opus 5 |
| Max output | 128K tokens | — | — |
| Input price / 1M | $5 ↗ | — | — |
| Output price / 1M | $25 ↗ | — | — |
| Cached input / 1M | $0.5 ↗ | — | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $30 | — | — |
| Modalities | text · vision | text · vision | Tie |
| Released | 2026-07-24 | 2025-07-09 | — |
| Cited benchmark scores | 41 | 17 | — |
Reliability
Provider status
Live from /statusMore matchups
Claude Opus 5 vs …
Models sharing the most benchmarksMore matchups
Grok 4 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Claude Opus 5 and Grok 4 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.