DeepSeek API pricing: every model, every rate

DeepSeek publishes different rates for peak and off-peak hours, and discounts cache hits heavily. That makes a single headline number misleading, so below is the standard rate for each model.

Full rate card

40 billable models in total, 33 of them text models. On those text models: 0 carry a rate verified against the vendor's own published pricing, the rest are gateway rates only. USD per 1M tokens, checked 2026-09-16.

Output is billed at a multiple of input on most models. The Out/In column is that multiple — it matters more than the input price once output dominates your bill.

Across 33 priced text models the blended rate spans $0.362375 (deepseek-v3.2) to $18 (deepseek-v4-pro-202606) per 1M tokens — a 50× spread. Picking the wrong tier is usually the single most expensive mistake here.

What DeepSeek costs at three usage levels

Priced on deepseek-v4-flash at $0.22 in / $0.66 out; deepseek-v4-pro at $0.66 in / $1.98 out per 1M tokens, with a 5:1 input-to-output ratio — roughly what an interactive workload produces.

The first column is this vendor's lowest-cost tier; the second is the model closest to its median rate, which is nearer what a real deployment bills. Swap in your own token split in the calculator.

Same budget, other vendors

3 models from other vendors priced within 35% of deepseek-v4-pro on a blended rate — the cheapest, the dearest and one in between. This is the horizontal check a single-vendor pricing page cannot give you.

Blended comparison (input + output per 1M tokens). Price is one axis — these are the models worth testing side by side at this budget, not interchangeable substitutes.

Why the output rate decides the bill

Across the 33 priced DeepSeek text models, the output rate runs from 1.5× to 4.0× the input rate, with a median of 3.0×. Output tokens are what a chat or agent turn produces — on a typical workload they are the smaller half of the token count but the larger half of the bill.

Worked out: at a 3.0× multiple, a job whose token count is 80% input still spends 43% of its cost on the output it generates. Comparing vendors on the input rate alone hides that.

What DeepSeek is good at

DeepSeek is the default choice when output volume dominates — R1-class reasoning at a fraction of US frontier pricing.

Detailed pages for DeepSeek models

The models we cover in depth, each with worked cost at three usage levels and a comparison against similarly priced alternatives.

• DeepSeek V4 Flash — $0.22 / $0.66 per 1M tokens
• DeepSeek V4 Pro — $0.66 / $1.98 per 1M tokens

Other vendors

Every rate card here is built the same way, from the same price snapshot (2026-09-16), so the numbers are comparable across vendors:

Anthropic · OpenAI · Google · Alibaba · xAI · Zhipu · Moonshot · Meta · MiniMax · ByteDance

FAQ

Create free account

See all 827 models → or use the cost calculator with your own token split.