wehub-resource-sync
267 lines
5.7 KiB
TOML
267 lines
5.7 KiB
TOML
name = "GPT-5.5"
|
|
description = "Default frontier GPT for coding, computer use, research, and knowledge work"
|
|
family = "gpt"
|
|
release_date = "2026-04-23"
|
|
last_updated = "2026-04-23"
|
|
attachment = true
|
|
reasoning = true
|
|
temperature = false
|
|
tool_call = true
|
|
structured_output = true
|
|
knowledge = "2025-12-01"
|
|
open_weights = false
|
|
|
|
[limit]
|
|
context = 1_050_000
|
|
input = 922_000
|
|
output = 128_000
|
|
|
|
[modalities]
|
|
input = ["text", "image", "pdf"]
|
|
output = ["text"]
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Bench Pro"
|
|
score = 58.6
|
|
metric = "resolve rate"
|
|
source = "https://www.anthropic.com/news/claude-opus-4-8"
|
|
date = "2026-05-28"
|
|
|
|
[[benchmarks]]
|
|
name = "Terminal-Bench"
|
|
score = 78.2
|
|
metric = "success rate"
|
|
harness = "Terminus-2"
|
|
version = "2.1"
|
|
source = "https://www.anthropic.com/news/claude-opus-4-8"
|
|
date = "2026-05-28"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Atlas Codebase QnA"
|
|
score = 45.43
|
|
metric = "score"
|
|
harness = "Codex"
|
|
source = "https://labs.scale.com/leaderboard/sweatlas-qna"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Atlas Refactoring"
|
|
score = 44.79
|
|
metric = "score"
|
|
harness = "Codex"
|
|
source = "https://labs.scale.com/leaderboard/sweatlas-refactoring"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Atlas Test Writing"
|
|
score = 42.59
|
|
metric = "score"
|
|
harness = "Codex"
|
|
source = "https://labs.scale.com/leaderboard/sweatlas-tw"
|
|
|
|
[[benchmarks]]
|
|
name = "Artificial Analysis Coding Agent Index"
|
|
score = 65.3
|
|
metric = "average pass@1"
|
|
harness = "Codex"
|
|
variant = "xhigh"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Atlas Codebase QnA"
|
|
score = 80.8
|
|
metric = "pass@1"
|
|
harness = "Codex"
|
|
variant = "xhigh"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Bench Pro"
|
|
score = 30.9
|
|
metric = "pass@1"
|
|
harness = "Codex"
|
|
variant = "xhigh"
|
|
dataset = "hard-aa"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "Terminal-Bench"
|
|
score = 84.1
|
|
metric = "pass@1"
|
|
harness = "Codex"
|
|
variant = "xhigh"
|
|
version = "2.1"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "Artificial Analysis Coding Agent Index"
|
|
score = 60.4
|
|
metric = "average pass@1"
|
|
harness = "Codex"
|
|
variant = "medium"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Atlas Codebase QnA"
|
|
score = 79.1
|
|
metric = "pass@1"
|
|
harness = "Codex"
|
|
variant = "medium"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Bench Pro"
|
|
score = 26.2
|
|
metric = "pass@1"
|
|
harness = "Codex"
|
|
variant = "medium"
|
|
dataset = "hard-aa"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "Terminal-Bench"
|
|
score = 75.8
|
|
metric = "pass@1"
|
|
harness = "Codex"
|
|
variant = "medium"
|
|
version = "2.1"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "Artificial Analysis Coding Agent Index"
|
|
score = 57.8
|
|
metric = "average pass@1"
|
|
harness = "Cursor CLI"
|
|
variant = "medium"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Atlas Codebase QnA"
|
|
score = 75
|
|
metric = "pass@1"
|
|
harness = "Cursor CLI"
|
|
variant = "medium"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "SWE-Bench Pro"
|
|
score = 24.9
|
|
metric = "pass@1"
|
|
harness = "Cursor CLI"
|
|
variant = "medium"
|
|
dataset = "hard-aa"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "Terminal-Bench"
|
|
score = 73.4
|
|
metric = "pass@1"
|
|
harness = "Cursor CLI"
|
|
variant = "medium"
|
|
version = "2.1"
|
|
source = "https://artificialanalysis.ai/agents/coding-agents"
|
|
|
|
[[benchmarks]]
|
|
name = "Terminal-Bench"
|
|
score = 82.7
|
|
metric = "success rate"
|
|
version = "2.0"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "GPQA Diamond"
|
|
score = 93.6
|
|
metric = "accuracy"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "Humanity's Last Exam"
|
|
score = 41.4
|
|
metric = "accuracy"
|
|
variant = "no tools"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "Humanity's Last Exam"
|
|
score = 52.2
|
|
metric = "accuracy"
|
|
variant = "with tools"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "OSWorld-Verified"
|
|
score = 78.7
|
|
metric = "success rate"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "BrowseComp"
|
|
score = 84.4
|
|
metric = "accuracy"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "MMMU Pro"
|
|
score = 81.2
|
|
metric = "accuracy"
|
|
variant = "no tools"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "ARC-AGI-2"
|
|
score = 85.0
|
|
metric = "accuracy"
|
|
variant = "Verified"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "FrontierMath"
|
|
score = 51.7
|
|
metric = "accuracy"
|
|
dataset = "Tier 1-3"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "FrontierMath"
|
|
score = 35.4
|
|
metric = "accuracy"
|
|
dataset = "Tier 4"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "GDPval"
|
|
score = 84.9
|
|
metric = "wins or ties"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "MCP Atlas"
|
|
score = 75.3
|
|
metric = "success rate"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "Toolathlon"
|
|
score = 55.6
|
|
metric = "success rate"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|
|
|
|
[[benchmarks]]
|
|
name = "τ²-Bench Telecom"
|
|
score = 98.0
|
|
metric = "success rate"
|
|
variant = "original prompts"
|
|
source = "https://openai.com/index/introducing-gpt-5-5/"
|
|
date = "2026-04-23"
|