Claude Fable 5

+ contribute

Claude Fable 5 is a closed-weights language model from Anthropic, released June 2026.

81.4±0.5#11 of 719 LLMsfrom 47 benchmark results across 11 sources

$ how it compares

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

#	model	score	where	released
9	Gemini 4 Argon (limited preview) Google	82±1	limited preview	2026-09
10	GPT-5.6 Sol OpenAI	82±1	api $8.00/M	2026-07
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
13	GPT-5.4 Pro OpenAI	81±1	api $67.50/M	2026-03
	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▁▁▃▆█▆▅▅▅▄▃▃ #8 of 300 · 32% of its score · 14 tests

Beats GPT-5.5 Pro, Gemini 4 Argon (limited preview), GPT-5.6 Sol here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Terminal-Bench 4.0 +1	44.5% ✎	█▅▅▅▅·▅▃▃▅▅▅ beats 14/20	very hard · avg 0.5%	▇▇▇ 6.7%	sharp · at its level · several runs
Arena Agent (October 2026 aggregate IPS)	0.0821 IPS ✎	▁▁▁▄▂▃▇█▂▃▁▃ beats 43/49	avg -0.312 IPS	▇▇ 4.4%	sharp
SWE-bench Pro (split unspecified) lab	80.0% ✎	▄█·▄·▄██···▄ beats 6/9	very hard · avg 7.6%	▇▇ 4.3%	sharp
SWE-bench Multilingual lab	86.6% ✎	▃··▃·▅▃▃█··▃ beats 4/9	hard · avg 31%	▇ 2.1%	typical
τ³-bench banking_knowledge (v1.0.1 grading)	39.7% ✎	▃▃▂▃·▆▃▂█▅▅▂ beats 18/25	very hard · avg 4.1%	▇ 1.8%	at its level
SWE-bench Verified lab	95.0% ✎	▁▁▁▁▁▁▂▄█▆▁▃ beats 130/131	hard · avg 28%	▇ 1.8%	sharp · maxed out
Chartography with tools lab	84.2% ✎	▃········▃██ beats 3/7	very hard · avg 4.7%	▇ 1.8%	sharp
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@1	65.3% ✎	▃··▅··▃█▃▅▃▅ beats 11/12	very hard · avg 3.1%	▇ 1.7%	sharp · at its level

6 more tests

test	result	vs the others in its run	difficulty	counts	why
Terminal-Bench 2.1 lab	84.3% ✎	▁·▁·▁▁▄▃▄█▃▇ beats 44/56	hard · avg 17%	▇ 1.4%	typical
Humanity's Last Exam with tools lab +1	63.8% ✎	▁······▁·▄·█ beats 5/10	very hard · avg 14%	▇ 1.3%	at its level
OSWorld-Verified lab	85.0% ✎	█▄·▂▆▂▄▂▆▄▆█ beats 24/25	hard · avg 29%	▇ 1.2%	typical
DeepSWE v1.1 (resolved) lab	69.7% ✎	█·····████·█ beats 3/5	hard · avg 18%	▇ 1.2%	at its level
Toolathlon (Tool Decathlon) lab	61.7% ✎	█···█··█·███ beats 5/5	hard · avg 16%	▇ 1.2%	at its level
BrowseComp with context management lab	87.4% ✎	▅·▅▃·█▃·█▅·▃ beats 13/14	medium · avg 44%	▇ 0.9%	typical

coding▁▁▁▃▆██▇▆▄▃▁ #14 of 458 · 16% of its score · 7 tests

Beats GPT-5.6 Sol here, though they rank higher overall. Trails Qwen3.8 Max (0902), Kimi K3, Muse Spark 1.3 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Pro (split unspecified) lab	80.0% ✎	▄█·▄·▄██···▄ beats 6/9	very hard · avg 7.6%	▇▇ 4.3%	sharp
Arena WebDev (Code Arena)	1625.3 ✎	▁▂▄▃▅█▄▇▆▄▁▁ beats 103/119	avg 981	▇▇ 3.9%	sharp
SWE-bench Multilingual lab	86.6% ✎	▃··▃·▅▃▃█··▃ beats 4/9	hard · avg 31%	▇ 2.1%	typical
SWE-bench Verified lab	95.0% ✎	▁▁▁▁▁▁▂▄█▆▁▃ beats 130/131	hard · avg 28%	▇ 1.8%	sharp · maxed out
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@1	65.3% ✎	▃··▅··▃█▃▅▃▅ beats 11/12	very hard · avg 3.1%	▇ 1.7%	sharp · at its level
DeepSWE v1.1 (resolved) lab	69.7% ✎	█·····████·█ beats 3/5	hard · avg 18%	▇ 1.2%	at its level
Luau Leetcode fully solved (Torpedo Software)	85.1% ✎	▁▁▁▁·▁▂▃▅▃█▇ beats 46/49	hard · avg 18%	▇ 1.0%	typical

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #2 of 393 * · 2.2% of its score · 1 test

Beats GPT-6 Astra, Claude Opus 5.5, GPT-6.1 Sol here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1504.3 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 377/379	avg 1268	▇ 2.2%	blunt

math▁▂▃▅▆▅▄█▄▁▃▁ #6 of 398 · 31% of its score · 11 tests

Beats Claude Sonnet 5.5, Claude Opus 5, GPT-5.5 Pro here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
FrontierMath tier 4 v2	90.2% ✎	▁▂▁▂█▅▂▂▃▁▁▁ beats 60/63	very hard · avg 0.1%	▇▇▇ 8.6%	sharp · at its level
FrontierMath tiers 1–3 v2	87.0% ✎	▂··▁▁▃▄▄█▅▃▃ beats 74/82	very hard · avg 1.2%	▇▇ 5.5%	sharp
MathArena ArXivMath April 2026 (40 problems)	72.5% ✎	▁·····▅██▄·▁ beats 18/20	very hard · avg 7.2%	▇▇ 4.1%	sharp · at its level
MathArena ArXivMath June 2026 (48 problems)	85.4% ✎	▂···▄·██▆▆▄▂ beats 16/19	very hard · avg 4.6%	▇▇ 4.0%	sharp
MathArena BrokenArXiv April 2026 (61 problems)	54.1% ✎	▂▂▆█▃▃▂▂▂··▂ beats 14/18	very hard · avg 0.7%	▇ 2.2%	sharp · at its level · noisy
MathArena ArXivLean June 2026 (46 problems)	15.2% ✎	▄····▄▄███·▄ beats 3/9	very hard · avg 0.7%	▇ 2.0%	sharp · at its level
MathArena BrokenArXiv May 2026 (50 problems)	44.5% ✎	██·▄▄▄····▄▄ beats 7/8	very hard · avg 1.6%	▇ 1.6%	sharp · at its level
MathArena BrokenArXiv June 2026 (54 problems)	47.8% ✎	▄··█▂▂▂▆▄··▂ beats 9/14	very hard · avg 1.1%	▇ 1.5%	sharp · at its level · noisy

3 more tests

test	result	vs the others in its run	difficulty	counts	why
MathArena ArXivMath May 2026 (40 problems)	87.5% ✎	▁··▁▄█▄▁·▃·▁ beats 16/17	very hard · avg 10%	▇ 1.4%	at its level
FrontierMath: Erdős problems	0.0% ✎	█··········▆ beats 2/6	very hard · avg 0.0%	▇ 0.3%	sharp · too hard for it · noisy
OTIS mock AIME 2024–2025	99.7% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 189/193	very hard · avg 8.5%	▇ 0.1%	sharp · maxed out · noisy

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #3 of 147 · 3.4% of its score · 4 tests

Beats GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5.1 here, though they rank higher overall. Trails Qwen3.8 Max (2.4T-A95B), Claude Mythos Preview (limited preview) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Arena vision (style controlled)	1308.8 ✎	▁▁▂▂▂▂▄▅▅▆██ beats 145/145	avg 1068	▇ 1.5%	blunt
CharXiv (reasoning questions)	86.6% ✎	▂▅▃▅▅▆▆▅▆█▆▃ beats 98/100	hard · avg 33%	▇ 0.8%	small run
Arena document (style controlled)	1498.9 ✎	▁▁▃▃▄▃▇█▅▇▄▅ beats 37/37	avg 1335	▇ 0.7%	blunt
CharXiv descriptive questions	94.3% ✎	▂▄▄▅▇▄█▇▇▂▃▂ beats 56/58	easy · avg 66%	▇ 0.4%	maxed out · small run

web search▂▁▂▆▄█▃▃▁▁·▁ #19 of 80 · 3.1% of its score · 3 tests

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	63.8% ✎	▁······▁·▄·█ beats 5/10	very hard · avg 14%	▇ 1.3%	at its level
BrowseComp with context management lab	87.4% ✎	▅·▅▃·█▃·█▅·▃ beats 13/14	medium · avg 44%	▇ 0.9%	typical
Arena search systems (style controlled)	1228.9 ✎	▁·▁▁▂▂▂█▅▂▃▁ beats 29/30	avg 1055	▇ 0.9%	blunt

other tests13 not in a task · 25% of its score

test	result	vs the others in its run	difficulty	counts	why
EBR-bench (Epoch)	39.5% ✎	▄▂·▆█▂▆▄▆▆·▄ beats 16/23	very hard · avg 0.3%	▇▇▇ 6.3%	sharp · at its level
Mystery Game Puzzles (Epoch)	52.0% ✎	█··▂▃▇▄█▂▃▁▁ beats 65/74	very hard · avg 9.4%	▇▇ 4.3%	sharp · at its level · noisy
Furniture Assembly (Epoch)	35.8% ✎	█▆▆▆▆▂▅▂▂▂·▅ beats 13/30	very hard · avg 1.2%	▇▇ 4.1%	sharp · at its level
Humanity's Last Exam lab +1	57.8% ✎	▂·····▂·▄▂█▄ beats 8/10	very hard · avg 5.8%	▇ 2.7%	at its level
Chess Puzzles (Epoch)	41.0% ✎	█▁▁▁▁▂▁▄▂▃▁▁ beats 121/135	very hard · avg 5.9%	▇ 2.3%	sharp · at its level · noisy
Global-MMLU lab	93.6% ✎	▄▄····▄▄··▄█ beats 4/6	medium · avg 63%	▇ 1.1%	typical
SimpleQA Verified	70.7% ✎	▁▁▃▂▃▃▆▆█▃▂▅ beats 73/78	very hard · avg 8.0%	▇ 0.8%	at its level
GPQA Diamond	85.9% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 148/208	hard · avg 40%	▇ 0.8%	sharp · maxed out

5 more tests

test	result	vs the others in its run	difficulty	counts	why
AutomationBench public set (Zapier README)	46.2% ✎	▄······▄█··▄ beats 2/4	very hard · avg 10%	▇ 0.8%	at its level
HealthBench lab	61.2% ✎	█·▄·····▄▄▄▄ beats 2/6	hard · avg 29%	▇ 0.6%	blunt · at its level
MirrorCode (Epoch)	63.9% ✎	█·▄▄·▄·▄·▄·█ beats 6/8	very hard · avg 1.5%	▇ 0.5%	sharp · at its level · noisy
RobloxQA v2.0 multiple choice (Torpedo Software)	97.0% ✎	▁▁▃▁█▄▃▃▆▃▂▂ beats 41/49	easy · avg 78%	▇ 0.4%	maxed out
NTCIR SciClaimEval Subtask 1 pair-accuracy (dev, single model)	74.4% ✎	▃▃···▃·▃▃█·▃ beats 8/8	medium · avg 45%	▇ 0.3%	blunt · at its level

$ where to run it

Through an API

api · $20.00/M · $10 in / $50 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… 34%, Anthropic system cards and mo… 25%, MathArena ArXivMath / BrokenA… 17% and 8 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	11	▇▇▇▇▇▇▇▇▇▇ 34%	CC BY 4.0
Anthropic system cards and model announcements (lab-reported scores) · lab-reported	16	▇▇▇▇▇▇▇▇ 25%	Factual vendor-reported score (not licensed content)
MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets	7	▇▇▇▇▇ 17%	CC BY-SA 4.0
LMArena leaderboard dataset	6	▇▇▇▇ 14%	CC BY 4.0
Terminal-Bench 4.0 official leaderboard submissions	1	▇ 3%	Apache-2.0
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
Meridian Labs inspect_charxiv evaluation report (GitHub README)	2	▇ 1%	MIT
Zapier AutomationBench README: public-set pass rates	1	▇ 1%	MIT
SciTrue: Reliable Scientific Claim Validation with Frontier and Open Language Models at the NTCIR SciClaimEval Task — authors' own single-model dev evaluation	1	▇ 0%	CC BY 4.0
prior · pull toward the middle		▇ 0%
weighted average		= 81.4