Atlas

Models

← All models

K2-V2-Instruct

LLM360 · 2025-12-05 · 72.6B parameters

K2-V2-Instruct is LLM360's instruction-tuned 70B K2-V2 model exposing low, medium, and high reasoning-effort settings through the chat template. It inherits K2-V2's 14.969T pretraining and mid-training tokens and adds three supervised fine-tuning epochs of approximately 17.4B tokens each.

Benchmark scores

BenchmarkScore
AA-LCR33.3
AA-Omniscience Index-48.1
AIME 202578.3
Artificial Analysis Intelligence Index14.2
Artificial Analysis Omniscience Accuracy18.3
Artificial Analysis Omniscience Hallucination Rate76.4
Artificial Analysis Openness Index88.9
Capability133.2
CritPt0.0
Global MMLU Lite78.6
GPQA Diamond68.1
Humanity's Last Exam (Text-Only)9.8
IFBench (Artificial Analysis)60.1
LiveCodeBench (Artificial Analysis)69.4
MMLU Pro78.6
SciCode (Artificial Analysis)28.6
Terminal-Bench Hard9.8
τ²-bench (Telecom)27.8
Loading Atlas data…