Alexey Shurov.
Free tool

Claude vs DeepSeek API pricing, compared model by model

On 25 September 2026 the cheapest Claude model costs $0.25 per million input tokens against $0.0482 for the cheapest DeepSeek model, and the typical DeepSeek model is the cheaper of the two.

Prices checked 25 September 2026 at 06.00 UTC . Source OpenRouter public model list . US dollars per million tokens

At a glance

Claude (Anthropic)DeepSeek (DeepSeek)
Models listed1515
Cheapest modelClaude 3 Haiku, $0.25 in, $1.25 outDeepSeek V4 Flash 0423, $0.0482 in, $0.0963 out
Most expensive modelClaude Opus 4.1, $15.00 in, $75.00 outR1, $0.70 in, $2.50 out
Median input price$5.00$0.27
Median output price$25.00$0.89
Largest context1M, Claude Opus 5.51.31072M, DeepSeek V4 Flash 0731
Newest modelClaude Opus 5.5, 2026-09-22DeepSeek V4.1 Flash, 2026-09-10
Models with a batch price131

Every model, with your monthly cost

ModelProviderInputOutputCached inputBatch in / outContextYour monthListed
Claude Opus 5.5Anthropic$4.00$20.00$0.20$2.00 / $10.001M$540.002026-09-22
DeepSeek V4.1 FlashDeepSeek$0.075$0.30$0.0015$0.112 / $0.3361.04858M$9.002026-09-10
Claude Fable 5.1Anthropic$10.00$50.00$0.25$5.00 / $25.001M$1,350.002026-09-01
DeepSeek V4 Flash Vision ExpDeepSeek$0.22$0.66$0.0071.04858M$23.102026-08-21
DeepSeek V4 Pro 0813DeepSeek$0.3458$1.04$0.01151.04858M$36.312026-08-12
DeepSeek V4 Flash 0731DeepSeek$0.03$0.32$0.0161.31072M$6.602026-07-31
Claude Opus 5Anthropic$5.00$25.00$0.50$2.50 / $12.501M$675.002026-07-24
Claude Sonnet 5Anthropic$2.00$10.00$0.20$1.00 / $5.001M$270.002026-06-30
Claude Fable 5Anthropic$10.00$50.00$1.00$5.00 / $25.001M$1,350.002026-06-09
Claude Opus 4.8Anthropic$5.00$25.00$0.50$2.50 / $12.501M$675.002026-05-27
DeepSeek V4 Pro 0423DeepSeek$0.7129$1.43$0.05941.04858M$64.162026-04-24
DeepSeek V4 Flash 0423DeepSeek$0.0482$0.0963$0.00961.04858M$4.332026-04-24
Claude Opus 4.7Anthropic$5.00$25.00$0.50$2.50 / $12.501M$675.002026-04-16
Claude Sonnet 4.6Anthropic$3.00$15.00$0.30$1.50 / $7.501M$405.002026-02-17
Claude Opus 4.6Anthropic$5.00$25.00$0.50$2.50 / $12.501M$675.002026-02-04
DeepSeek V3.2
retiring 2026-09-28
DeepSeek$0.269$0.40$0.1345164K$22.142025-12-01
Claude Opus 4.5Anthropic$5.00$25.00$0.50$2.50 / $12.50200K$675.002025-11-24
Claude Haiku 4.5Anthropic$1.00$5.00$0.10$0.50 / $2.50200K$135.002025-10-15
Claude Sonnet 4.5Anthropic$3.00$15.00$0.30$1.50 / $7.501M$405.002025-09-29
DeepSeek V3.2 Exp
retiring 2026-09-28
DeepSeek$0.27$0.41164K$22.352025-09-29
DeepSeek V3.1 Terminus
retiring 2026-09-28
DeepSeek$0.27$1.00$0.135164K$31.202025-09-22
DeepSeek V3.1DeepSeek$0.25$0.95$0.13164K$29.252025-08-21
Claude Opus 4.1Anthropic$15.00$75.00$1.50$7.50 / $37.50200K$2,025.002025-08-05
R1 0528DeepSeek$0.50$2.15$0.35164K$62.252025-05-28
Claude Sonnet 4Anthropic$3.00$15.00$0.30200K$405.002025-05-22
DeepSeek V3 0324DeepSeek$0.25$1.00164K$30.002025-03-24
R1 Distill Llama 70B
retiring 2026-09-28
DeepSeek$0.80$0.808K$60.002025-01-23
R1DeepSeek$0.70$2.5064K$79.502025-01-20
DeepSeek V3DeepSeek$0.32$0.89164K$32.552024-12-26
Claude 3 HaikuAnthropic$0.25$1.25$0.03200K$33.752024-03-13

All Claude prices . All DeepSeek prices . Every provider

Common questions

Is Claude cheaper than DeepSeek?

On 25 September 2026, the median Claude model on this list costs $5.00 per million input tokens and $25.00 per million output tokens, against $0.27 and $0.89 for DeepSeek. By that measure DeepSeek is cheaper, but compare the specific models you would use, since both providers sell models at very different price points.

What is the cheapest Claude model compared with the cheapest DeepSeek model?

The cheapest Claude model is Claude 3 Haiku at $0.25 in and $1.25 out per million tokens. The cheapest DeepSeek model is DeepSeek V4 Flash 0423 at $0.0482 in and $0.0963 out.

Which has the larger context window, Claude or DeepSeek?

The largest Claude context on this list is 1M tokens (Claude Opus 5.5) and the largest DeepSeek context is 1.31072M tokens (DeepSeek V4 Flash 0731).

How should I choose between Claude and DeepSeek for a production agent?

Price per token is one input. Run both on a set of your own real cases, count how often each needs a retry or a human check, and compare the total cost of a correct answer rather than the cost of a token.

Choosing a model for a production agent

Price is one input. Quality on your own cases, latency and the cost of retries decide the rest. Tell me what you are building.

shurco.aiGuidesFree toolsSolutionsInsightsRSSllms.txt