DeepSeek has 7 models tracked on this site, from DeepSeek V3 in 2024 to DeepSeek V4.1 Flash, released Sep 10, 2026. Its highest-ranked model, DeepSeek V4.1 Flash, is #14 of 39 on the Singularity Index with a score of 47.5.
Newest first. Bars show each model's Singularity Index score out of 100. Provisional scores have partial benchmark coverage and are not ranked yet.
Native vision model using a causal encoder-decoder MoE architecture, with 8B active parameters for input and 16B for output. Replaces V4 Flash API aliases; V4 Pro service continues under updated September 14 guidance. Listed prices are peak rates; off-peak rates are half.
Vision-capable efficiency tier of V4, pairing GPQA 91.3 with multimodal coverage at open-weight cost.
Post-trained update to DeepSeek V4 Flash with stronger reasoning, coding, and agentic performance.
Flagship successor to V3 (Max Effort reasoning). Open-weight, competitive with closed frontier at a fraction of the cost. Awaiting Multimodal eval coverage.
Matched GPT-5 class at a fraction of the cost, open-weight MoE
Open-weight reasoning model that triggered the DeepSeek market shock
Trained for $5.5M, proved frontier performance was possible at low cost