Atlas

Models

← All models

Apriel-1.6-15B-Thinker

ServiceNow AI Research · 2025-12-09 · 14.9B parameters

Apriel-1.6-15B-Thinker is ServiceNow AI Research's December 2025 update to the Pixtral-derived Apriel-1.5 multimodal reasoning checkpoint. It improves text and image reasoning while reducing reasoning-token usage by more than 30% relative to V1.5, supports tool use, and remains deployable on a single GPU. The checkpoint is released under MIT.

Benchmark scores

BenchmarkScore
AA-LCR50.3
AA-Omniscience Index-58.5
AIME 202588.0
Artificial Analysis Intelligence Index20.5
Artificial Analysis Omniscience Accuracy16.9
Artificial Analysis Omniscience Hallucination Rate90.9
Artificial Analysis Openness Index50.0
Capability142.3
CritPt0.3
GPQA Diamond73.3
Humanity's Last Exam (Text-Only)9.8
IFBench (Artificial Analysis)69.1
LiveCodeBench (Artificial Analysis)80.7
MMLU Pro79.0
SciCode (Artificial Analysis)37.3
Terminal-Bench Hard14.4
τ²-bench (Telecom)69.3
Loading Atlas data…