Frontier Log

Benchmark · Inactive · No longer reported

GSM8K

Created by OpenAI, 27 Oct 2021 · Higher is better · Website

Results

ModelScoreSourceSource check
DeepSeek-V4.1-Flash-101717 Oct 2026 · DeepSeek95.4%DeepSeek abliterated refuses nothing & is just as intelligent as the regular modellm-eval-harness; thinking off; 1,319 problems; paired with Morph's abliterated copy; run by Morph, not chartedSource image
CheckedA person checked this number against its source on 25 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
DeepSeek V4.1 Flash 1017 Abliterated (Morph)17 Oct 2026 · DeepSeek95.1%DeepSeek abliterated refuses nothing & is just as intelligent as the regular modellm-eval-harness; thinking off; 1,319 problems; run by Morph, not chartedSource image
CheckedA person checked this number against its source on 25 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.

All benchmarks