DeepSeek
DeepSeek is an AI model provider.Tokenando tracks 27 DeepSeek models, with input pricing from $0.070/M and an average blended cost of $0.551/M. Its flagship model is DeepSeek V4 Pro 0423.
Open-weights frontier from a Chinese hedge-fund-backed lab. R1 reset price/performance expectations for reasoning models.
MODELS TRACKED
27
3 categories
FLAGSHIP
DeepSeek V4 Pro 0423
text->text
MIN INPUT
$0.070/M
cheapest model in family
AVG BLENDED
$0.551/M
across 27 priced models
MAX CONTEXT
1,048,576
largest window in family
Reasoning
10 models
R1 Distill Qwen 32Bprofile
text->text · 32,768 ctx
in $0.290/Mout $0.290/M
tokenizer: Qwen · cron:openrouter
DeepSeek-R1profile
Live API · 64,000 ctx
in $0.550/Mout $2.190/M
Chain-of-thought · manual-seed
DeepSeek-R1-0528profile
Live API · 64,000 ctx
in $0.550/Mout $2.190/M
Latest R1 · manual-seed
DeepSeek-R1-Distill-70Bprofile
Live API · 128,000 ctx
in $0.350/Mout $0.880/M
Distilled reasoning · manual-seed
DeepSeek-R1-Distill-32Bprofile
Live API · 128,000 ctx
in $0.200/Mout $0.500/M
32B distilled · manual-seed
DeepSeek-R1-Distill-8Bprofile
Live API · 128,000 ctx
in $0.070/Mout $0.200/M
8B distilled · manual-seed
DeepSeek-R1-Distill-14Bprofile
Live API · 128,000 ctx
in $0.100/Mout $0.350/M
14B distilled · manual-seed
R1 Distill Llama 70Bprofile
text->text · 131,072 ctx
in $0.800/Mout $0.800/M
tokenizer: Llama3 · cron:openrouter
R1 0528profile
text->text · 163,840 ctx
in $0.500/Mout $2.150/M
R1profile
text->text · 64,000 ctx
in $0.700/Mout $2.500/M
Multimodal
3 models
DeepSeek V4 Flash Vision Exp
text+image->text · 1,048,576 ctx
in $0.440/Mout $1.320/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V4.1 Flash
text+image->text · 1,048,576 ctx
in $0.300/Mout $1.200/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V4.1 Flash (batch)
text+image->text · 1,048,576 ctx
in $0.112/Mout $0.336/M
Efficient
14 models
DeepSeek V3profile
text->text · 163,840 ctx
in $0.257/Mout $1.029/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V3 0324profile
text->text · 163,840 ctx
in $0.290/Mout $1.140/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek-Coder-V2profile
Live API · 128,000 ctx
in $0.140/Mout $0.280/M
Code specialist · manual-seed
DeepSeek V3.1 Terminusprofile
text->text · 163,840 ctx
in $0.300/Mout $1.000/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V3.2 Expprofile
text->text · 163,840 ctx
in $0.270/Mout $0.410/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V3.2 Specialeprofile
text->text · 163,840 ctx
in $0.400/Mout $1.200/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V4 Flashprofile
text->text · 1,048,576 ctx
in $0.140/Mout $0.280/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V3.1profile
text->text · 32,768 ctx
in $0.150/Mout $0.750/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V4 Flash 0731
text->text · 1,048,576 ctx
in $0.140/Mout $0.280/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V4 Pro
text->text · 1,048,576 ctx
in $0.435/Mout $0.870/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V4 Pro 0813
text->text · 1,048,576 ctx
in $0.435/Mout $0.870/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V3.2profile
text->text · 131,072 ctx
in $0.214/Mout $0.322/M
tokenizer: DeepSeek · cron:openrouter
DeepSeek V4 Pro 0423
text->text · 1,048,576 ctx
in $1.600/Mout $3.200/M
Price change pending review (input 1.6 → 0.9483, output 3.2 → 1.8966, cached input 0.135 → 0.079025, ≥30% move) — showing last verified price, not the unreviewed new one.
DeepSeek V4 Flash 0423
text->text · 1,048,576 ctx
in $0.087/Mout $0.174/M
Price change pending review (input 0.08708 → 0.0173, output 0.17416 → 1.28, ≥30% move) — showing last verified price, not the unreviewed new one.
Frequently Asked Questions
How many models does DeepSeek offer?
Tokenando tracks 27 DeepSeek models.
How much do DeepSeek models cost?
DeepSeek model input pricing starts at $0.070 per million tokens, with an average blended cost of $0.551 per million across the 27 priced models we track.
What is DeepSeek's flagship model?
DeepSeek's flagship model is DeepSeek V4 Pro 0423. It is the highest-tier DeepSeek model we track, with input pricing of $1.600 per million tokens.
What model categories does DeepSeek cover?
DeepSeek covers 3 categories: reasoning, multimodal and efficient.