whatmodel.sh / Claude Opus 4.5

Claude Opus 4.5

+ contribute

Claude Opus 4.5 is a closed-weights language model from Anthropic, released November 2025.

72.1±0.6#65 of 719 LLMsfrom 39 benchmark results across 19 sources

$ how it compares

The LLMs ranked either side of Claude Opus 4.5, out of 828 with a score, then more from Anthropic.

#	model	score	where	RAM	released
68	Gemini 3 Flash (Preview) Google	72±1	api $1.13/M		2025-12
69	DeepSeek V4 Pro (Preview, Apr 2026) DeepSeek	72±1	local	~604GB	2026-04
70	Claude Opus 4.5 Anthropic	72±1	api $10.00/M		2025-11
71	Qwen3.7 Plus Alibaba	72±1	api $0.70/M		2026-06
72	Kimi K2.7 Code Moonshot AI	72±1	api $1.71/M +local	~359GB	2026-06
	more from Anthropic
2	Claude Opus 5.5 Anthropic	87±1	api $8.00/M		2026-09
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

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #54 of 300 · 27% of its score · 12 tests

Beats Gemini 3.1 Pro (Preview), GPT-5.2, Gemini 3.6 Flash here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Verified	76.7% ✎	▁··▁·▁▂▁▄█▄▁ beats 21/29	hard · avg 19%	▇▇ 6.2%	sharp · at its level
MCP-Atlas (public set) +1	65.2% ✎	▄······▄▄▄▄█ beats 3/6	hard · avg 25%	▇▇ 4.3%	at its level · several runs
Terminal-Bench 2.0 with Terminus 2 agent (benchmark authors' runs)	57.8% ✎	▃··▃·▃▃██▅▃▅ beats 14/14	very hard · avg 5.8%	▇ 3.4%	sharp · at its level
τ²-bench aggregate	91.6% ✎	▄····█▄▄···▄ beats 5/5	medium · avg 40%	▇ 2.2%	typical
τ³-bench banking_knowledge (v1.0.1 grading)	24.7% ✎	▃▃▂▃·▆▃▂█▅▅▂ beats 7/25	very hard · avg 4.1%	▇ 2.1%	at its level
SWE-bench Pro public set, Qwen-refined	57.1% ✎	▄······▄█·▄▄ beats 5/5	hard · avg 18%	▇ 1.8%	at its level
τ³-bench airline (tau2-bench ≥ v1.0.1)	84.0% ✎	▃······▃·▃█▃ beats 6/6	medium · avg 49%	▇ 1.5%	typical
BFCL V4 (December 2025 archive)	77.5% ✎	▁▁▁▂██▅▅▆▂▂▃ beats 77/77	hard · avg 32%	▇ 1.5%	at its level

4 more tests

test	result	vs the others in its run	difficulty	counts	why
τ³-bench retail (tau2-bench ≥ v1.0.1)	79.6% ✎	▄▄·█··▄·▄··▄ beats 4/6	medium · avg 46%	▇ 1.3%	typical
τ²-bench Telecom lab	98.2% ✎	▁·▅█▆▅▂▃▃▂▃▁ beats 84/89	hard · avg 29%	▇ 1.2%	sharp · noisy
τ³-bench telecom (tau2-bench ≥ v1.0.1)	92.3% ✎	█·▄█▄······▄ beats 5/6	medium · avg 62%	▇ 1.1%	typical
τ²-Bench-Verified (GPT-5.1 user)	82.0% ✎	▂··▄···▂▄▄█▄ beats 12/13	medium · avg 45%	▇ 0.5%	typical

coding▁▁▁▃▆██▇▆▄▃▁ #54 of 458 · 15% of its score · 4 tests

Beats GPT-5.4, DeepSeek V4 Flash Vision (Exp), Gemini 3.1 Pro (Preview) here, though they rank higher overall. Trails GLM-5.1, Mistral Large 4 (public preview), Qwen3.7 Plus here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Verified	76.7% ✎	▁··▁·▁▂▁▄█▄▁ beats 21/29	hard · avg 19%	▇▇ 6.2%	sharp · at its level
Arena WebDev (Code Arena)	1493.3 ✎	▁▂▄▃▅█▄▇▆▄▁▁ beats 68/119	avg 981	▇▇ 5.7%	sharp
SWE-bench Pro public set, Qwen-refined	57.1% ✎	▄······▄█·▄▄ beats 5/5	hard · avg 18%	▇ 1.8%	at its level
WebApp1K-Duo-React pass@1 (onekq-ai)	76.3% ✎	▁··▁·▁▁▁▆█▇█ beats 41/45	hard · avg 32%	▇ 1.1%	typical

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #29 of 393 * · 3.2% of its score · 1 test

Beats Claude Sonnet 5.5, GPT-6 Sol, GPT-5.6 Terra here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1473.6 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 348/379	avg 1268	▇ 3.2%	blunt

math▁▂▃▅▆▅▄█▄▁▁▁ #95 of 398 · 27% of its score · 5 tests

Beats Claude Sonnet 4.6 here, though they rank higher overall. Trails Gemini 2.5 Flash, GLM-4.5-Air, Granite 4.2 30B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
FrontierMath tiers 1–3 v2	34.4% ✎	▂··▁▁▃▄▄█▅▃▃ beats 23/82	very hard · avg 1.2%	▇▇▇▇▇▇ 16%	sharp · at its level
FrontierMath tier 4 v2	4.9% ✎	▁▂▃▂█▅▂▂▁▁▁▁ beats 8/63	very hard · avg 0.1%	▇▇▇ 8.5%	sharp
MATH-Vision	74.3% ✎	█·····▄▄·▄·▄ beats 0/5	medium · avg 46%	▇ 1.2%	typical
OTIS mock AIME 2024–2025	86.1% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 127/193	very hard · avg 8.5%	▇ 1.1%	sharp · noisy
FormalProofBench (200 graduate-level Lean 4 problems, agentic)	33.5% ✎	▃▅·▃··▅▃█··▃ beats 10/10	very hard · avg 3.0%	▇ 0.7%	noisy

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #57 of 147 · 4.6% of its score · 3 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-5 here, though they rank higher overall. Trails Gemma 4 26B-A4B, Qwen3.6 Flash, Qwen3.5 35B-A3B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
CharXiv (reasoning questions) lab	65.7% ✎	███··█·██··█ beats 0/6	hard · avg 29%	▇ 2.3%	at its level
MATH-Vision	74.3% ✎	█·····▄▄·▄·▄ beats 0/5	medium · avg 46%	▇ 1.2%	typical
Arena document (style controlled)	1464.6 ✎	▁▁▃▃▄▃▇█▅▇▄▅ beats 21/37	avg 1335	▇ 1.1%	blunt

web search▂▁▂▆▄█▃▃▁▁·▁ #37 of 80 * · 1.3% of its score · 1 test

Beats Gemini 3.1 Pro (Preview), Qwen3.7 Max, Gemini 3.5 Flash here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), Llama 4 Scout (109B-A17B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Arena search systems (style controlled)	1187.2 ✎	▁·▁▁▂▂▂█▅▂▁▁ beats 12/30	avg 1055	▇ 1.3%	blunt

other tests15 not in a task · 31% of its score

test	result	vs the others in its run	difficulty	counts	why
Furniture Assembly (Epoch)	28.3% ✎	█▆▆▆▆▂▅▂▂▂·▅ beats 7/30	very hard · avg 1.2%	▇▇ 4.4%	sharp
EBR-bench (Epoch)	14.3% ✎	▄▂·▆█▂▆▄▆▆·▄ beats 9/23	very hard · avg 0.3%	▇▇ 4.3%	sharp
ARC-AGI-2 public evaluation set (ARC Prize model_baseline, pass@2)	40.3% ✎	█··▂▅▁▃▁▃·▁▃ beats 20/24	very hard · avg 0.3%	▇ 3.4%	sharp · at its level · noisy
GPQA Diamond	86.0% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 151/208	hard · avg 40%	▇ 3.3%	sharp
Mystery Game Puzzles (Epoch)	22.0% ✎	█··▂▃▇▄█▂▃▁▁ beats 38/74	very hard · avg 9.4%	▇ 2.8%	sharp · noisy
Chess Puzzles (Epoch)	12.0% ✎	█▁▁▁▁▃▁▄▂▂▁▁ beats 59/135	very hard · avg 5.9%	▇ 2.1%	sharp · noisy
ARC-AGI-1 public evaluation set (ARC Prize model_baseline, pass@2)	86.6% ✎	▁·▇▃▁·█▇▃·▇▃ beats 20/27	very hard · avg 4.1%	▇ 2.0%	sharp · noisy
EQ-Bench 3 (emotional intelligence, judge Elo)	1308.4 ✎	▁·▂▁▄▅▅▃█▄▁▁ beats 50/62	avg 1044	▇ 2.0%	typical

7 more tests

test	result	vs the others in its run	difficulty	counts	why
NL2Repo	43.2% ✎	▄·····█·▄▄·▄ beats 5/5	very hard · avg 8.2%	▇ 1.8%	at its level
SimpleQA Verified	45.7% ✎	▁▁▃▂▃▃▆▆█▃▂▅ beats 42/78	very hard · avg 8.0%	▇ 1.2%	at its level
AraGen-v3 3C3H score (Inception, Claude Sonnet 4 judge)	80.3% ✎	▁▁·▁▂▂▄█▄▅▄▄ beats 62/66	hard · avg 31%	▇ 1.2%	typical
HLE-Verified full set, text-only (Qwen Team runs)	38.8% ✎	█·█·····█··█ beats 1/3	very hard · avg 14%	▇ 1.1%	at its level
SOOHAK Challenge (340 problems, Avg@3)	10.4% ✎	█·█████··█·█ beats 3/7	very hard · avg 4.1%	▇ 0.6%	typical
SOOHAK-Mini (702 problems, Avg@3)	51.4% ✎	▄··▄·▄·▄▄█·▄ beats 1/7	medium · avg 36%	▇ 0.4%	blunt · at its level
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)	89.1% ✎	▁▂▁▂▅█▂▃▄▃▁▁ beats 33/98	very easy · avg 90%	▇ 0.0%	blunt

$ where to run it

Through an API

api · $10.00/M · $5 in / $25 out per 1M tokens · 200k 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… 50%, LMArena leaderboard dataset 11%, τ²/τ³-bench leaderboard: Sier… 6% and 16 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	10	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 50%	CC BY 4.0
LMArena leaderboard dataset	4	▇▇▇ 11%	CC BY 4.0
τ²/τ³-bench leaderboard: Sierra's own evaluation runs	4	▇▇ 6%	MIT
ARC Prize Foundation model_baseline outputs on ARC-AGI-1/2 public evaluation sets	2	▇▇ 5%	MIT
Qwen/Qwen3.6-27B model card	2	▇ 4%	apache-2.0
Anthropic system cards and model announcements (lab-reported scores) · lab-reported	2	▇ 3%	Factual vendor-reported score (not licensed content)
Qwen/Qwen3.5-397B-A17B model card	2	▇ 3%	apache-2.0
Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces — benchmark authors' own agent-model trials	1	▇ 3%	CC BY 4.0
zai-org/GLM-5 model card	1	▇ 2%	MIT
GLM-5 Team (Zhipu AI) — GLM-5 technical report	1	▇ 2%	CC BY 4.0
EQ-Bench 3 canonical leaderboard results	1	▇ 2%	MIT
Berkeley Function Calling Leaderboard result archive	1	▇ 1%	Apache-2.0
Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval)	1	▇ 1%	Apache-2.0
WebApp1K models leaderboard (onekq-ai)	1	▇ 1%	MIT
HLE-Verified: A Systematic Verification and Structured Revision of Humanity's Last Exam — Qwen Team re-evaluation on revised Humanity's Last Exam	1	▇ 1%	CC BY 4.0
Soohak: A Mathematician-Curated Benchmark for Evaluating Research-level Math Capabilities of LLMs — authors' own model runs	2	▇ 1%	CC BY-SA 4.0
FormalProofBench (Vals AI, arXiv:2603.26996v1)	1	▇ 1%	CC-BY-4.0
Amazon AGI — τ²-Bench-Verified	1	▇ 1%	MIT
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 72.1