name = "Claude Opus 5"
description = "Strongest Claude Opus model for coding, agents, and professional work"
family = "claude-opus"
release_date = "2026-07-24"
last_updated = "2026-07-24"
attachment = true
reasoning = true
temperature = false
tool_call = true
open_weights = false
knowledge = "2026-05"

[limit]
context = 1_000_000
output = 128_000

[modalities]
input = ["text", "image", "pdf"]
output = ["text"]

[[benchmarks]]
name = "SWE-Bench Verified"
score = 96.0
metric = "resolved"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "SWE-Bench Pro"
score = 79.2
metric = "resolve rate"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "SWE-Bench Multilingual"
score = 89.5
metric = "resolve rate"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "SWE-Bench Multimodal"
score = 59.4
metric = "resolve rate"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "DeepSWE"
score = 68.8
metric = "resolve rate"
version = "1.1"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "FrontierCode"
score = 53.4
metric = "mean@5"
variant = "medium effort"
dataset = "Main"
version = "1.1"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "Frontier-Bench"
score = 43.3
metric = "mean reward"
variant = "max effort"
harness = "mini-SWE-agent"
version = "v0.1"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "BrowseComp"
score = 90.8
metric = "accuracy"
variant = "single agent"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "Humanity's Last Exam"
score = 56.3
metric = "accuracy"
variant = "no tools"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "Humanity's Last Exam"
score = 64.7
metric = "accuracy"
variant = "with tools"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "DeepSearchQA"
score = 95.0
metric = "F1"
variant = "max effort"
source = "https://www.anthropic.com/news/claude-opus-5"
date = "2026-07-24"

[[benchmarks]]
name = "OSWorld"
score = 70.6
metric = "success rate"
version = "2.0"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "GDPval-AA"
score = 1861
metric = "Elo"
variant = "max effort"
version = "v2"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "AA-Briefcase"
score = 1720
metric = "Elo"
variant = "max effort"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "AutomationBench"
score = 26.0
metric = "success rate"
variant = "max effort"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "ARC-AGI-1"
score = 97.5
metric = "accuracy"
variant = "max effort"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "ARC-AGI-2"
score = 90.4
metric = "accuracy"
variant = "max effort"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "ARC-AGI-3"
score = 30.2
metric = "RHAE"
variant = "high effort"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"

[[benchmarks]]
name = "HealthBench Professional"
score = 59.8
metric = "score"
variant = "max effort"
source = "https://www.anthropic.com/claude-opus-5-system-card"
date = "2026-07-24"