Solar Mini 4 scores 24 on the Artificial Analysis Intelligence Index but costs about five times GPT-6 Luna (max) per task

Upstage released Solar Mini 4, a proprietary reasoning model it says has 3B active parameters—a size Artificial Analysis cannot independently verify. Artificial Analysis scores it 24 on its Intelligence Index, 16 points above Solar Pro 3. It measured $0.36 per task versus $0.07 for GPT-6 Luna (max); 48% of Solar Mini 4’s repeated context was served from cache, compared with 99% for Luna.