Compare · ModelsLive · 2 picked · head to head
GPT-5.5 vs Claude Mythos Preview
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Claude Mythos Preview wins on 2/3 benchmarks
Claude Mythos Preview wins 2 of 3 shared benchmarks. Leads in knowledge · agentic.
Category leads
knowledge·Claude Mythos Previewagentic·Claude Mythos Previewcoding·GPT-5.5
Hype vs Reality
Attention vs performance
GPT-5.5
#2 by perf·no signal
Claude Mythos Preview
#4 by perf·#2 by attention
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
Anthropic
$380.0B·Tier 1
Head to head
3 benchmarks · 2 models
GPT-5.5Claude Mythos Preview
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.
GPT-5.5
93.6
Claude Mythos Preview
94.5
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.
GPT-5.5
78.7
Claude Mythos Preview
79.6
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.
GPT-5.5
82.7
Claude Mythos Preview
82.0
Full benchmark table
| Benchmark | GPT-5.5 | Claude Mythos Preview |
|---|---|---|
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 93.6 | 94.5 |
OSWorld OSWorld · tests AI agents on real-world computer tasks across operating systems, including web browsing, file management, and application use. | 78.7 | 79.6 |
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. | 82.7 | 82.0 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $5.00 | $30.00 | 400K tokens (~200 books) | $112.50 | |
| — | — | 1.0M tokens (~500 books) | — |