name = "Claude Opus 4.6"
description = "High-end Claude for difficult coding, planning, and slower expert reasoning"
family = "claude-opus"
release_date = "2026-02-05"
last_updated = "2026-03-13"
attachment = true
reasoning = true
temperature = true
tool_call = true
knowledge = "2025-05-31"
open_weights = false

[limit]
context = 1_000_000
output = 128_000

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

[[benchmarks]]
name = "SWE-Bench Pro"
score = 51.9
metric = "resolve rate"
dataset = "public"
source = "https://labs.scale.com/leaderboard/swe_bench_pro_public"

[[benchmarks]]
name = "SWE-Atlas Codebase QnA"
score = 33.3
metric = "score"
harness = "Claude Code"
source = "https://labs.scale.com/leaderboard/sweatlas-qna"

[[benchmarks]]
name = "SWE-Atlas Codebase QnA"
score = 30
metric = "score"
harness = "Mini-SWE-Agent"
source = "https://labs.scale.com/leaderboard/sweatlas-qna"

[[benchmarks]]
name = "SWE-Atlas Refactoring"
score = 35.58
metric = "score"
harness = "Claude Code"
source = "https://labs.scale.com/leaderboard/sweatlas-refactoring"

[[benchmarks]]
name = "SWE-Atlas Test Writing"
score = 36.67
metric = "score"
harness = "Claude Code"
source = "https://labs.scale.com/leaderboard/sweatlas-tw"

[[benchmarks]]
name = "SWE-Atlas Test Writing"
score = 36.08
metric = "score"
harness = "Mini-SWE-Agent"
source = "https://labs.scale.com/leaderboard/sweatlas-tw"

[[benchmarks]]
name = "Artificial Analysis Coding Agent Index"
score = 51.3
metric = "average pass@1"
harness = "Claude Code"
variant = "medium"
source = "https://artificialanalysis.ai/agents/coding-agents"

[[benchmarks]]
name = "SWE-Atlas Codebase QnA"
score = 71.9
metric = "pass@1"
harness = "Claude Code"
variant = "medium"
source = "https://artificialanalysis.ai/agents/coding-agents"

[[benchmarks]]
name = "SWE-Bench Pro"
score = 11.8
metric = "pass@1"
harness = "Claude Code"
variant = "medium"
dataset = "hard-aa"
source = "https://artificialanalysis.ai/agents/coding-agents"

[[benchmarks]]
name = "Terminal-Bench"
score = 70.2
metric = "pass@1"
harness = "Claude Code"
variant = "medium"
version = "2.1"
source = "https://artificialanalysis.ai/agents/coding-agents"