Atlas

Benchmarks

← All benchmarks

Codeforces - DeepSeek-reported expected rating

Code

Expected Codeforces rating computed by DeepSeek's internal Codeforces evaluation (problems from 10 recent Division 2 contests graded with expert-crafted test cases, as described in the DeepSeek-R1 report). The contest set and rating procedure are DeepSeek-internal and not comparable with Codeforces leaderboard Elo, so the row is diagnostic.

Top models (higher is better)

ModelScore
DeepSeek-R12029
o1-mini1820
DeepSeek-R1-Distill-Qwen-32B1691
DeepSeek-R1-Distill-Llama-70B1633
DeepSeek R1 Distill Qwen 14B1481
DeepSeek-R1-Zero1444
QwQ-32B-Preview1316
DeepSeek R1 Distill Llama 8B1205
DeepSeek-R1-Distill-Qwen-7B1189
DeepSeek-V31134
DeepSeek-R1-Distill-Qwen-1.5B954
GPT-4o (2024-05-13)759
Claude 3.5 Sonnet (Oct 2024)717
Loading Atlas data…