whatmodel.sh / Claude Mythos Preview (limited preview)

Claude Mythos Preview (limited preview)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Claude Mythos Preview (limited preview) is a closed-weights language model from Anthropic, released April 2026. Licence: proprietary.

Limited preview · broad public access is not yet available. Access announcement ↗

Catalogue identity notes

Anthropic's gated research-preview model announced April 7, 2026 with Project Glasswing ('a general-purpose, unreleased frontier model'); access limited to launch partners and vetted organisations, and Anthropic says 'We do not plan to make Claude Mythos Preview generally available'. Distinct from Claude Mythos 5 / Mythos 5.1, which Anthropic documents as the same models as Claude Fable 5 / Fable 5.1 with more permissive safeguards (platform.claude.com mythos-5 and mythos-5-1 pages; Fable 5.1 system card p.1). Comparison identity only: no public run route is listed (the $25/$125 per MTok post-preview pricing on the Glasswing page is prospective). Precedent: gemini-4-argon (cyber-gated limited preview). Scores are Anthropic's own (system cards).

Model information sources

Release date ↗

81.3±1.1#12 of 719 LLMsfrom 12 benchmark results across 1 source

$ how it compares

The LLMs ranked either side of Claude Mythos Preview (limited preview), out of 828 with a score, then more from Anthropic.

#modelscorewhereRAMreleased
10GPT-5.6 Sol OpenAI82±1api $8.00/M2026-07
11Claude Fable 5 Anthropic81±1api $20.00/M2026-06
12Claude Mythos Preview (limited preview) Anthropic81±1limited preview2026-04
13GPT-5.4 Pro OpenAI81±1api $67.50/M2026-03
14GPT-5.5 OpenAI80±1api $11.25/M2026-04
more from Anthropic
2Claude Opus 5.5 Anthropic87±1api $8.00/M2026-09
4Claude Fable 5.1 Anthropic85±1api $20.00/M2026-09
5Claude Sonnet 5.5 Anthropic84±1api $4.00/M2026-09

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #10 of 300 · 67% of its score · 7 tests

Beats GPT-5.5 Pro, GPT-5.6 Sol here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Pro (split unspecified) lab77.8% ✎▄█·▄·▄██···▄ beats 4/9very hard · avg 7.6%▇▇▇▇▇▇▇▇▇ 23%sharp
SWE-bench Multilingual lab87.3% ✎▃··▃·▅▃▃█··▃ beats 5/9hard · avg 31%▇▇▇▇ 11%typical
SWE-bench Verified lab93.9% ✎▁▁▁▁▁▁▂▄█▆▁▃ beats 129/131hard · avg 28%▇▇▇▇ 10%sharp · maxed out
Humanity's Last Exam with tools lab64.7% ✎▁······▁·▄·█ beats 8/10very hard · avg 14%▇▇▇ 7.0%at its level
OSWorld-Verified lab85.4% ✎█▄·▂▆▂▄▂▆▄▆█ beats 25/25hard · avg 29%▇▇▇ 6.5%typical
Terminal-Bench 2.0 lab82.0% ✎▂▂▅▅████▆▂▃▃ beats 35/36very hard · avg 12%▇▇ 5.6%small run
BrowseComp with context management lab86.9% ✎▅·▅▃·█▃·█▅·▃ beats 12/14medium · avg 44%▇▇ 4.8%typical
coding▁▁▁▃▆██▇▆▄▃▁ #9 of 458 · 43% of its score · 3 tests

Beats GPT-5.6 Sol, Claude Fable 5 here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Pro (split unspecified) lab77.8% ✎▄█·▄·▄██···▄ beats 4/9very hard · avg 7.6%▇▇▇▇▇▇▇▇▇ 23%sharp
SWE-bench Multilingual lab87.3% ✎▃··▃·▅▃▃█··▃ beats 5/9hard · avg 31%▇▇▇▇ 11%typical
SWE-bench Verified lab93.9% ✎▁▁▁▁▁▁▂▄█▆▁▃ beats 129/131hard · avg 28%▇▇▇▇ 10%sharp · maxed out
math▁▂▃▅▆▅▄█▄▁▃▁ #9 of 398 * · 2.6% of its score · 1 test

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

testresultvs the others in its rundifficultycountswhy
USAMO 2026 lab97.6% ✎█·██······██ beats 4/4hard · avg 16%▇ 2.6%sharp · noisy
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #1 of 147 * · 5.5% of its score · 1 test

Beats GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5.1 here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
CharXiv (reasoning questions) lab86.2% ✎███··█·██··█ beats 6/6hard · avg 29%▇▇ 5.5%typical
web search▂▁▂▆▄█▃▃▁▁·▁ #20 of 80 * · 12% of its score · 2 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.

testresultvs the others in its rundifficultycountswhy
Humanity's Last Exam with tools lab64.7% ✎▁······▁·▄·█ beats 8/10very hard · avg 14%▇▇▇ 7.0%at its level
BrowseComp with context management lab86.9% ✎▅·▅▃·█▃·█▅·▃ beats 12/14medium · avg 44%▇▇ 4.8%typical
other tests3 not in a task · 24% of its score
testresultvs the others in its rundifficultycountswhy
Humanity's Last Exam lab56.8% ✎▂·····▂·▄▂█▄ beats 6/10very hard · avg 5.8%▇▇▇▇▇▇ 14%at its level
GPQA Diamond lab94.5% ✎▁▁▁▁▂▄▅▅███▄ beats 304/310hard · avg 40%▇▇▇ 6.8%sharp · maxed out · small run
MMMLU lab92.7% ✎▂▁▁▂▃▂▄▄▅██▃ beats 46/46medium · avg 61%▇ 3.2%maxed out · small run

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · Anthropic system cards and mo… 99%

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

sourceresultsshare of the scorelicence
Anthropic system cards and model announcements (lab-reported scores) · lab-reported12▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 99%Factual vendor-reported score (not licensed content)
prior · pull toward the middle▇ 1%
weighted average= 81.3

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json