Skip to content
ModelCap

DeepSeek

5 across 1 category board · 8 superseded · deepseek.com

Language ModelCap Index

5
Language rankModelGravityContextInput tokensOutput tokensReleased
#19DeepSeek V4.1 FlashOpen weights87.91M$0.30$1.2012 days ago
#25DeepSeek V4 Pro 0813Open weights61.41M$1.32$3.96last month
#60DeepSeek V3.2Open weights57.0164K$0.269$0.409 months ago
#61R1 0528Open weights32.0164K$0.50$2.15last year
#69DeepSeek V3.1 TerminusOpen weights28.2164K$0.27$1.0012 months ago

Superseded versions

8
Board statusModelGravityContextInput tokensOutput tokensReleased
Superseded by DeepSeek V4.1 FlashDeepSeek V4 Flash 0731Open weights
Language
85.71.3M$0.04$0.64last month
Superseded by DeepSeek V4 Pro 0813DeepSeek V4 Pro 0423Open weights
Language
66.51M$0.955$1.915 months ago
Superseded by DeepSeek V4.1 FlashDeepSeek V4 Flash 0423Open weights
Language
76.31M$0.089$0.1775 months ago
Superseded by DeepSeek V3.2DeepSeek V3.2 ExpOpen weights
Language
47.8164K$0.27$0.4111 months ago
Superseded by DeepSeek V3.2DeepSeek V3.1Open weights
Language
49.5164K$0.25$0.95last year
Superseded by DeepSeek V3.2DeepSeek V3 0324Open weights
Language
50.9164K$0.25$1.00last year
Superseded by R1 0528R1Open weights
Language
44.364K$0.70$2.50last year
Superseded by DeepSeek V3.2DeepSeek V3License unverified
Language
41.7164K$0.32$0.89last year

Other off-board models

2
Board statusModelGravityContextInput tokensOutput tokensReleased
Off boardDeepSeek V4 Flash Vision ExpOpen weights
Language
51.01M$0.22$0.66last month
Off boardR1 Distill Llama 70BOpen weights
Language
12.48K$0.80$0.80last year

DeepSeek models: common questions

How many AI models has DeepSeek released?

ModelCap tracks 15 canonical DeepSeek models as of 22 September 2026: 15 language models (5 with a public ModelCap Index position). Superseded versions and equivalent serving routes stay listed on this page, marked as such.

What is the best DeepSeek model?

DeepSeek V4.1 Flash is DeepSeek's highest-ranked model as of 22 September 2026, at ModelCap Index position #19 with a score of 77.8, followed by DeepSeek V4 Pro 0813 (#25). The table lists every ranked DeepSeek model in Index order.

What is the cheapest DeepSeek model?

DeepSeek V4 Flash 0423 at $0.177 per 1M output tokens ($0.089 per 1M input) as of 22 September 2026, the lowest listed price among DeepSeek's 15 priced language models. Prices are the lowest current provider offer on the OpenRouter catalogue.

Which DeepSeek model has the largest context window?

DeepSeek V4 Flash 0731, with a published context window of 1.3M tokens. Individual providers can serve less than the published maximum; each model page lists every endpoint's own limit.

Are DeepSeek models open-weight?

14 of 15 tracked DeepSeek models publish downloadable weights (open weights). ModelCap labels weight availability from repository metadata, not whether the training stack is open source; the licence class is shown beside each row.

What is the newest DeepSeek model?

DeepSeek V4.1 Flash, first listed on 10 September 2026 and currently ModelCap Index #19. New DeepSeek models appear here as soon as they pass automatic identity checks; the launches page tracks releases across every publisher.

Explore further