Head-to-head comparison across 3 benchmark categories
Claude Mythos Preview
84
GPT-5.2
81
Pick Claude Mythos Preview if you want the stronger benchmark profile. GPT-5.2 only becomes the better choice if you want the cheaper token bill.
Agentic
+25.9 difference
Coding
+28.2 difference
Multimodal
+13.2 difference
Claude Mythos Preview
GPT-5.2
$25 / $125
$2 / $8
N/A
73 t/s
N/A
130.34s
1M
400K
Pick Claude Mythos Preview if you want the stronger benchmark profile. GPT-5.2 only becomes the better choice if you want the cheaper token bill.
Claude Mythos Preview has the cleaner overall profile here, landing at 84 versus 81. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
Claude Mythos Preview's sharpest advantage is in coding, where it averages 83.8 against 55.6. The single biggest benchmark swing on the page is OSWorld-Verified, 79.6% to 47.3%.
Claude Mythos Preview is also the more expensive model on tokens at $25.00 input / $125.00 output per 1M tokens, versus $2.00 input / $8.00 output per 1M tokens for GPT-5.2. That is roughly 15.6x on output cost alone. Claude Mythos Preview gives you the larger context window at 1M, compared with 400K for GPT-5.2.
Claude Mythos Preview is ahead overall, 84 to 81. The biggest single separator in this matchup is OSWorld-Verified, where the scores are 79.6% and 47.3%.
Claude Mythos Preview has the edge for coding in this comparison, averaging 83.8 versus 55.6. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.
Claude Mythos Preview has the edge for agentic tasks in this comparison, averaging 80.9 versus 55. Inside this category, OSWorld-Verified is the benchmark that creates the most daylight between them.
Claude Mythos Preview has the edge for multimodal and grounded tasks in this comparison, averaging 92.7 versus 79.5. Inside this category, MMMU-Pro is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.