Claude Opus 5

+ contribute

Claude Opus 5 is a closed-weights language model from Anthropic, released July 2026.

82.7±0.5#6 of 719 LLMsfrom 48 benchmark results across 17 sources

$ how it compares

The LLMs ranked either side of Claude Opus 5, out of 828 with a score, then more from Anthropic.

#	model	score	where	released
4	Claude Fable 5.1 Anthropic	85±1	api $20.00/M	2026-09
5	Claude Sonnet 5.5 Anthropic	84±1	api $4.00/M	2026-09
6	Claude Opus 5 Anthropic	83±1	api $10.00/M	2026-07
7	GPT-6 Sol OpenAI	83±1	api $4.00/M	2026-09
8	GPT-5.5 Pro OpenAI	82±1	api $67.50/M	2026-04
	more from Anthropic
2	Claude Opus 5.5 Anthropic	87±1	api $8.00/M	2026-09
11	Claude Fable 5 Anthropic	81±1	api $20.00/M	2026-06
12	Claude Mythos Preview (limited preview) Anthropic	81±1	limited preview	2026-04

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▃ #6 of 300 · 36% of its score · 16 tests

test	result	vs the others in its run	difficulty	counts	why
Terminal-Bench 4.0 +1	51.8% ✎	█▅▅▅▅·▅▃▃▅▅▅ beats 16/20	very hard · avg 0.5%	▇▇▇ 7.5%	sharp · at its level · several runs
Arena Agent (October 2026 aggregate IPS)	0.0792 IPS ✎	▁▁▁▄▂▃▇█▂▃▁▃ beats 42/49	avg -0.312 IPS	▇▇ 4.8%	sharp
SWE-bench Pro (split unspecified) lab	79.2% ✎	▄█·▄·▄██···▄ beats 5/9	very hard · avg 7.6%	▇▇ 4.3%	sharp
GameLogicBench solve rate (72 Godot game-logic tasks, Claude Code scaffold)	52.8% ✎	█▄·█·▄·██▄·▄ beats 11/11	very hard · avg 4.7%	▇ 2.1%	at its level
SWE-bench Multilingual lab	89.5% ✎	▃··▃·▅▃▃█··▃ beats 7/9	hard · avg 31%	▇ 2.1%	typical
τ³-bench banking_knowledge (v1.0.1 grading)	48.7% ✎	▃▃▂▃·▆▃▂█▅▅▂ beats 24/25	very hard · avg 4.1%	▇ 2.0%	at its level
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@1	67.7% ✎	▃··▅··▃█▃▅▃▅ beats 12/12	very hard · avg 3.1%	▇ 1.9%	sharp · at its level
SWE-bench Verified lab	96.0% ✎	▁▁▁▁▁▁▂▄█▆▁▃ beats 131/131	hard · avg 28%	▇ 1.8%	sharp · maxed out

8 more tests

test	result	vs the others in its run	difficulty	counts	why
Chartography with tools lab +1	83.4% ✎	▃········▃██ beats 2/7	very hard · avg 4.7%	▇ 1.7%	sharp
Humanity's Last Exam with tools lab +1	63.6% ✎	▁······▁·▄·█ beats 6/10	very hard · avg 14%	▇ 1.4%	at its level
DeepSWE v1.1 (resolved) lab	68.8% ✎	█·····████·█ beats 2/5	hard · avg 18%	▇ 1.3%	at its level
Video2World V2WScore (coding agents reconstruct interactive simulated worlds from videos, OpenCode scaffold)	41.6% ✎	█··▄█·▄····▄ beats 6/6	very hard · avg 6.6%	▇ 1.3%	at its level
Thinkingbox-bench pass@1 (507 stateful business-workflow agent tasks, N=20)	66.5% ✎	▅··▃·▃·█▃▅██ beats 15/15	very hard · avg 4.9%	▇ 1.1%	at its level
BrowseComp with context management lab	90.8% ✎	▅·▅▃·█▃·█▅·▃ beats 14/14	medium · avg 44%	▇ 0.9%	typical
KlinikeBench strict pass@1 (333 interactive clinical encounters, Terminus-2)	29.4% ✎	▁▁·▁▁▁▁▅▃█▅█ beats 27/27	very hard · avg 2.4%	▇ 0.9%	typical
Toolathlon-Verified lab	80.6% ✎	▃·····█···▃▃ beats 5/5	hard · avg 35%	▇ 0.7%	blunt

coding▁▁▁▃▆██▇▆▄▃▃ #6 of 458 · 19% of its score · 8 tests

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Pro (split unspecified) lab	79.2% ✎	▄█·▄·▄██···▄ beats 5/9	very hard · avg 7.6%	▇▇ 4.3%	sharp
Arena WebDev (Code Arena)	1695.0 ✎	▁▂▄▃▅█▄▇▆▄▃▁ beats 114/119	avg 981	▇▇ 4.2%	sharp
GameLogicBench solve rate (72 Godot game-logic tasks, Claude Code scaffold)	52.8% ✎	█▄·█·▄·██▄·▄ beats 11/11	very hard · avg 4.7%	▇ 2.1%	at its level
SWE-bench Multilingual lab	89.5% ✎	▃··▃·▅▃▃█··▃ beats 7/9	hard · avg 31%	▇ 2.1%	typical
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@1	67.7% ✎	▃··▅··▃█▃▅▃▅ beats 12/12	very hard · avg 3.1%	▇ 1.9%	sharp · at its level
SWE-bench Verified lab	96.0% ✎	▁▁▁▁▁▁▂▄█▆▁▃ beats 131/131	hard · avg 28%	▇ 1.8%	sharp · maxed out
DeepSWE v1.1 (resolved) lab	68.8% ✎	█·····████·█ beats 2/5	hard · avg 18%	▇ 1.3%	at its level
Luau Leetcode fully solved (Torpedo Software)	82.2% ✎	▁▁▁▁·▁▂▃▅▃█▇ beats 38/49	hard · avg 18%	▇ 1.0%	typical

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #11 of 393 * · 2.4% of its score · 1 test

Beats GPT-6 Astra, GPT-6.1 Sol, Claude Sonnet 5.5 here, though they rank higher overall. Trails Claude Opus 4.6, Muse Spark 1.1, Claude Opus 4.7 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1489.5 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 369/379	avg 1268	▇ 2.4%	blunt

math▁▂▃▅▆▅▄█▄▁▃▁ #12 of 398 · 21% of its score · 6 tests

Trails GPT-5.5, Claude Mythos Preview (limited preview), Claude Fable 5 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
FrontierMath tier 4 v2	73.2% ✎	▁▂▁▂█▅▂▃▁▁▁▁ beats 54/63	very hard · avg 0.1%	▇▇▇ 8.0%	sharp
FrontierMath tiers 1–3 v2	85.6% ✎	▂··▁▁▃▄▄█▅▃▃ beats 72/82	very hard · avg 1.2%	▇▇ 5.0%	sharp
MathArena ArXivMath June 2026 (48 problems)	82.6% ✎	▂···▄·██▆▆▄▂ beats 14/19	very hard · avg 4.6%	▇▇ 3.9%	sharp
MathArena ArXivLean June 2026 (46 problems)	32.6% ✎	▄····▄▄███·▄ beats 7/9	very hard · avg 0.7%	▇ 2.3%	sharp · at its level
MathArena BrokenArXiv June 2026 (54 problems)	90.7% ✎	▄··█▂▂▂▆▄··▃ beats 14/14	very hard · avg 1.1%	▇ 1.6%	sharp · at its level · noisy
OTIS mock AIME 2024–2025	98.9% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 179/193	very hard · avg 8.5%	▇ 0.1%	sharp · maxed out · noisy

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #6 of 147 · 3.2% of its score · 4 tests

Beats GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5.1 here, though they rank higher overall. Trails Claude Opus 4.7, Gemini 3.7 Flash, Qwen3.8 Max (2.4T-A95B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Arena vision (style controlled)	1289.0 ✎	▁▁▂▂▂▂▄▅▅▆██ beats 134/145	avg 1068	▇ 1.6%	blunt
Arena document (style controlled)	1489.7 ✎	▁▁▃▃▄▃▇█▅▇▄▅ beats 33/37	avg 1335	▇ 0.8%	blunt
SciDocBench scientific-document workflows (overall, 496 instances)	62.6% ✎	▃▃···█▃▅▃·▃▅ beats 11/11	hard · avg 26%	▇ 0.8%	blunt · at its level
PureDocBench document parsing (Avg3 over Clean/Digital/Real)	77.1% ✎	▂·▁·▂▂▅█·▁▃▂ beats 19/22	medium · avg 62%	▇ 0.0%	blunt · at its level

web search▂▁▂▆▄█▃▃▁▁·▁ #16 of 80 * · 2.4% of its score · 2 tests

Beats Claude Fable 5.1, Claude Sonnet 5.5 here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), GPT-4o (2024-08-06) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Humanity's Last Exam with tools lab +1	63.6% ✎	▁······▁·▄·█ beats 6/10	very hard · avg 14%	▇ 1.4%	at its level
BrowseComp with context management lab	90.8% ✎	▅·▅▃·█▃·█▅·▃ beats 14/14	medium · avg 44%	▇ 0.9%	typical

other tests18 not in a task · 32% of its score

test	result	vs the others in its run	difficulty	counts	why
EBR-bench (Epoch)	45.7% ✎	▄▂·▆█▂▆▄▆▆·▄ beats 18/23	very hard · avg 0.3%	▇▇▇ 7.0%	sharp · at its level
Mystery Game Puzzles (Epoch)	59.0% ✎	█··▂▃▇▄█▂▃▁▁ beats 70/74	very hard · avg 9.4%	▇▇ 4.8%	sharp · at its level · noisy
Furniture Assembly (Epoch)	60.8% ✎	█▆▆▆▆▂▅▃▂▂·▅ beats 25/30	very hard · avg 1.2%	▇▇ 4.4%	sharp · at its level
Humanity's Last Exam lab +1	56.6% ✎	▂·····▂·▄▂█▄ beats 5/10	very hard · avg 5.8%	▇ 3.0%	at its level
Chess Puzzles (Epoch)	42.0% ✎	█▁▁▁▁▂▁▄▂▃▁▁ beats 122/135	very hard · avg 5.9%	▇ 2.5%	sharp · at its level · noisy
KMMLU-HARD (Daekeun Kim, zero-shot)	88.4% ✎	█···▁▁▁▂▃·▆▃ beats 23/23	very hard · avg 29%	▇ 1.7%	sharp
Global-MMLU lab	92.5% ✎	▄▄····▄▄··▄█ beats 3/6	medium · avg 63%	▇ 1.1%	typical
KoBALT-700 (Daekeun Kim)	88.1% ✎	▄▄█·····▄··▄ beats 5/5	hard · avg 37%	▇ 1.0%	typical

10 more tests

test	result	vs the others in its run	difficulty	counts	why
SimpleQA Verified	59.9% ✎	▁▁▃▂▃▃▆▆█▃▂▅ beats 63/78	very hard · avg 8.0%	▇ 0.9%	at its level
AutomationBench public set (Zapier README)	50.3% ✎	▄······▄█··▄ beats 4/4	very hard · avg 10%	▇ 0.8%	at its level
KMMLU-Pro (Daekeun Kim)	95.6% ✎	▄·▄█·····▄·▄ beats 5/5	medium · avg 40%	▇ 0.8%	typical
Icelandic LLM Leaderboard average (Miðeind, Sept 2026 suite)	68.6% ✎	▁▁▃▂▂▄▁▃▄▄▄█ beats 35/38	very hard · avg 14%	▇ 0.8%	at its level
GPQA Diamond	93.9% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 199/208	hard · avg 40%	▇ 0.8%	sharp · maxed out
CLIcK Korean (Daekeun Kim)	95.8% ✎	▆▄▄▄█▄▂·▄█·▄ beats 23/23	medium · avg 58%	▇ 0.8%	maxed out
HealthBench lab	67.1% ✎	█·▄·····▄▄▄▄ beats 4/6	hard · avg 29%	▇ 0.6%	blunt · at its level
MuSR Korean (Daekeun Kim)	86.5% ✎	█·····███·██ beats 5/5	medium · avg 42%	▇ 0.6%	typical
HAE-RAE 1.0 (Daekeun Kim)	95.8% ✎	▄▄▄▂▂█▂▄▂█▂▄ beats 22/22	medium · avg 63%	▇ 0.5%	maxed out
RobloxQA v2.0 multiple choice (Torpedo Software)	97.5% ✎	▁▁▃▁█▄▃▃▆▃▂▂ beats 42/49	easy · avg 78%	▇ 0.4%	maxed out

$ where to run it

Through an API

api · $10.00/M · $5 in / $25 out per 1M tokens · 1M context · docs.anthropic.com ↗ · price sourceStandard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ.

$ how the score was worked out · Epoch AI — internal evaluatio… 33%, Anthropic system cards and mo… 23%, LMArena leaderboard dataset 14% and 14 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
Epoch AI — internal evaluations	9	▇▇▇▇▇▇▇▇▇▇ 33%	CC BY 4.0
Anthropic system cards and model announcements (lab-reported scores) · lab-reported	15	▇▇▇▇▇▇▇ 23%	Factual vendor-reported score (not licensed content)
LMArena leaderboard dataset	5	▇▇▇▇ 14%	CC BY 4.0
MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets	3	▇▇ 8%	CC BY-SA 4.0
Daekeun Kim Korean LLM benchmark runs	6	▇▇ 5%	MIT
Terminal-Bench 4.0 official leaderboard submissions	1	▇ 4%	Apache-2.0
GameLogicBench: Evaluating Coding Agents on Runtime Game Logic with Tick-Level State Assertions (arXiv:2609.21562v2)	1	▇ 2%	CC BY 4.0
τ²/τ³-bench leaderboard: Sierra's own evaluation runs	1	▇ 2%	MIT
E2E-SWE: Benchmarking LLMs on Building Working Codebases from Scratch (arXiv:2609.38335v1)	1	▇ 2%	CC BY 4.0
Roblox LLM Leaderboard (Torpedo Software): RobloxQA v2.0 and Luau Leetcode	2	▇ 1%	MIT
Video2World: Benchmarking Coding Agents for Interactive World Modeling from Embodied Videos (arXiv:2610.04432v2)	1	▇ 1%	CC BY-SA 4.0
One Success Isn't Reliability: Thinkingbox, a Sandbox and Benchmark for Agents in Stateful Business Workflows (arXiv:2608.19741v4)	1	▇ 1%	CC BY 4.0
KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy (arXiv:2609.38480v2)	1	▇ 1%	CC BY 4.0
Zapier AutomationBench README: public-set pass rates	1	▇ 1%	MIT
Miðeind Icelandic LLM Leaderboard	1	▇ 1%	CC BY 4.0
SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding (arXiv:2609.05141v2)	1	▇ 1%	CC BY 4.0
How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)	1	▇ 0%	CC BY 4.0
prior · pull toward the middle		▇ 0%
weighted average		= 82.7