Inkling-Small

Benchmark Model Rank Results
instruction-following-on-ifbenchInkling-Small#1Accuracy: 82.2
mathematical-reasoning-on-hmmt-feb-2026Inkling-Small#4Accuracy: 90.2
physics-question-answering-on-critptInkling-Small#4Accuracy: 8.3
code-generation-on-scicodeInkling-Small#6Accuracy: 48.7
mathematical-reasoning-on-aime-2026Inkling-Small#6Accuracy: 95.5
code-generation-on-swe-bench-verifiedInkling-Small (Bash-only)#10Accuracy: 80.2
agents-on-toolathlon-verified-108-tasksInkling-Small#11Pass@1: 0.544
abstract-reasoning-on-arc-agiInkling-Small#13Accuracy: 84.0
abstract-reasoning-on-arc-agi-2Inkling-Small#13Accuracy: 40.1
multimodal-reasoning-on-mmmu-proInkling-Small#13Accuracy: 74.0
multiple-choice-question-answering-mcqa-on-gpqa-diamondInkling-Small#14Accuracy: 89.5
deep-research-agents-on-browsecompInkling-Small (with context management)#17Accuracy: 77.4
code-generation-on-swe-bench-proInkling-Small#24Resolved: 55.9
code-generation-on-terminal-bench-2-1Inkling-Small (Internal coding harness)#29Accuracy: 64.7