DeepSeek V4 Pro Climbs to No. 2 Among Open-Weight Models
DeepSeek has released DeepSeek V4 Pro and V4 Flash, its first new architecture since V3 and its first two-tier lineup, with Pro aimed at maximum capability and Flash at faster, lower-cost inference. On Artificial Analysis’s Intelligence Index, V4 Pro scored 52, up from 42 for V3.2, ranking second among open-weight reasoning models behind Kimi K2.6’s 54, while V4 Flash scored 47. Separate Vibe Code Benchmark results also placed DeepSeek V4 as the top open-weight model, ahead of Kimi K2.6 and Gemini 3.1 Pro.
Artificial Analysis said V4 Pro led open-weight models on the GDPval-AA real-world work benchmark with a score of 1,554, ahead of Kimi K2.6 at 1,484 and GLM-5.1 at 1,535. The model is DeepSeek’s largest yet at 1.6 trillion total parameters with 49 billion active parameters, versus 284 billion total and 13 billion active for Flash. Artificial Analysis estimated V4 Pro costs $1,071 to run its index, more than four times cheaper than Claude Opus 4.7’s $4,811 but still above several open-weight peers because of high token usage; Flash costs $113, and the firm reported hallucination rates of 94% for V4 Pro and 96% for Flash.
From the sources (5 posts)
@wesrothDeepSeek v4 has completely shattered the Vibe Code Benchmark (VCB), officially becoming the #1 open-weight model and the very first to cross the 40% threshold, landing just shy of 50%. The model decisively leaves the previous open-source l
@artificialanlysDeepSeek is back among the leading open weights models with the release of DeepSeek V4 Pro and V4 Flash, with V4 Pro second only to Kimi K2.6 on the Artificial Analysis Intelligence Index @deepseek_ai has released DeepSeek V4 Pro and V4 F
@artificialanlysDeepSeek V4 Pro scales DeepSeek’s architecture substantially, while V4 Flash is positioned for size efficiency: V4 Pro is DeepSeek’s largest model to date at 1.6T total parameters / 49B active, a major step up from the V3 family’s 671B tota
@artificialanlysDeepSeek V4 Pro leads open weights models on GDPval-AA, our agentic real-world work tasks benchmark. V4 Pro (Max) scores 1554, ahead of Kimi K2.6 (1484), GLM-5.1 (1535), GLM-5 (1402), and MiniMax-M2.7 (1514). V4 Flash (Reasoning, Max Effort
@artificialanlysLower cost than frontier models, but high token usage keeps costs above most open weights peers: DeepSeek V4 Pro costs $1,071 to run the Artificial Analysis Intelligence Index, more than 4x cheaper than Claude Opus 4.7 ($4,811) but above se