Compare · ModelsLive · 3 picked · head to head

Claude Mythos Preview vs GPT-5.5 vs Claude Opus 4.5

Side by side · benchmarks, pricing, and signals you can act on.

Winner summary

Claude Mythos Preview wins 4 of 6 shared benchmarks. Leads in knowledge · agentic.

Category leads
knowledge·Claude Mythos Previewagentic·Claude Mythos Previewcoding·GPT-5.5reasoning·GPT-5.5
Hype vs Reality
Claude Mythos Preview
#4 by perf·#2 by attention
DESERVED
GPT-5.5
#2 by perf·no signal
QUIET
Claude Opus 4.5
#113 by perf·no signal
QUIET
Best value
1.6x better value than Claude Opus 4.5
Claude Mythos Preview
no price
GPT-5.5
4.9 pts/$
$17.50/M
Claude Opus 4.5
3.0 pts/$
$15.00/M
Vendor risk
Anthropic logo
Anthropic
$380.0B·Tier 1
Medium risk
OpenAI logo
OpenAI
$840.0B·Tier 1
Medium risk
Anthropic logo
Anthropic
$380.0B·Tier 1
Medium risk
Head to head
Claude Mythos PreviewGPT-5.5Claude Opus 4.5
GPQA diamond
Claude Mythos Preview leads by +0.9
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude Mythos Preview
94.5
GPT-5.5
93.6
Claude Opus 4.5
81.4
OSWorld
Claude Mythos Preview leads by +0.9
OSWorld · tests AI agents on real-world computer tasks across operating systems, including web browsing, file management, and application use.
Claude Mythos Preview
79.6
GPT-5.5
78.7
Claude Opus 4.5
66.3
Terminal Bench
GPT-5.5 leads by +0.7
Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence.
Claude Mythos Preview
82.0
GPT-5.5
82.7
Claude Opus 4.5
63.1
ARC-AGI
GPT-5.5 leads by +15.0
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
GPT-5.5
95.0
Claude Opus 4.5
80.0
HLE
Claude Mythos Preview leads by +35.4
HLE (Humanity's Last Exam) · a reasoning benchmark designed to be the hardest public evaluation of AI. Questions span mathematics, physics, philosophy, and logic · curated to be at or beyond the frontier of human expert capability. Tested with and without tool augmentation. Claude Opus 4.7 scores 46.9% without tools and 54.7% with tools · making it one of the few benchmarks where the top score is below 60%.
Claude Mythos Preview
56.8
Claude Opus 4.5
21.4
SWE-Bench verified
Claude Mythos Preview leads by +17.3
SWE-bench Verified · 500 human-validated tasks from 12 real Python repositories (Django, Flask, scikit-learn, sympy, and others). Each task requires the model to produce a git patch that resolves a real GitHub issue and passes the test suite. The verified subset eliminates ambiguous tasks from the original SWE-bench. Claude Mythos Preview leads at 93.9%, crossing 90% for the first time in 2026. Opus 4.6 scores 80.8%. The benchmark remains the most-cited evaluation for code-generation capability.
Claude Mythos Preview
93.9
Claude Opus 4.5
76.7
Full benchmark table
BenchmarkClaude Mythos PreviewGPT-5.5Claude Opus 4.5
GPQA diamond
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
94.593.681.4
OSWorld
OSWorld · tests AI agents on real-world computer tasks across operating systems, including web browsing, file management, and application use.
79.678.766.3
Terminal Bench
Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence.
82.082.763.1
ARC-AGI
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
95.080.0
HLE
HLE (Humanity's Last Exam) · a reasoning benchmark designed to be the hardest public evaluation of AI. Questions span mathematics, physics, philosophy, and logic · curated to be at or beyond the frontier of human expert capability. Tested with and without tool augmentation. Claude Opus 4.7 scores 46.9% without tools and 54.7% with tools · making it one of the few benchmarks where the top score is below 60%.
56.821.4
SWE-Bench verified
SWE-bench Verified · 500 human-validated tasks from 12 real Python repositories (Django, Flask, scikit-learn, sympy, and others). Each task requires the model to produce a git patch that resolves a real GitHub issue and passes the test suite. The verified subset eliminates ambiguous tasks from the original SWE-bench. Claude Mythos Preview leads at 93.9%, crossing 90% for the first time in 2026. Opus 4.6 scores 80.8%. The benchmark remains the most-cited evaluation for code-generation capability.
93.976.7
Pricing · per 1M tokens · projected $/mo at 10M tokens
ModelInputOutputContextProjected $/mo
Anthropic logoClaude Mythos Preview1.0M tokens (~500 books)
OpenAI logoGPT-5.5$5.00$30.00400K tokens (~200 books)$112.50
Anthropic logoClaude Opus 4.5$5.00$25.00200K tokens (~100 books)$100.00