DeepSeek Pricing (2026)
Pure pay-as-you-go: you're billed per 1M tokens consumed, with no seats, no subscription floor, and no free tier or trial to ease you in. The twist is time-of-day pricing, where off-peak hours (outside 01:00-04:00 and 06:00-10:00 UTC on weekdays, plus Chinese holidays) cost half the peak rate. Search and Fetch APIs are a separate meter at $0.75 per 1,000 successful requests.
- deepseek-flash
- Contact Sales
- deepseek-v4-pro
- Contact Sales
Price watch
Get an email when DeepSeek changes its price.
No change recorded yet. You'll hear about the first one before most people notice.
One email per change, with the old and new price. No newsletter. Unsubscribe in one click.
deepseek-flash
- Json Output
- Tool Calls
- Responses API
- Anthropic API
- Chat Prefix Completion（Beta）
- FIM Completion（Beta）
- Vision
deepseek-v4-pro
- Json Output
- Tool Calls
- Responses API
- Anthropic API
- Chat Prefix Completion（Beta）
- FIM Completion（Beta）
Does DeepSeek have a free plan?
No. DeepSeek lists no free plan or trial on its pricing page. Pricing is custom; you'll need to talk to sales.
Want free? Anthropic, Baseten, and Braintrust have free plans: see all DeepSeek alternatives →
AI Pricing Analysis
Pricing Model
Pure pay-as-you-go: you're billed per 1M tokens consumed, with no seats, no subscription floor, and no free tier or trial to ease you in. The twist is time-of-day pricing, where off-peak hours (outside 01:00-04:00 and 06:00-10:00 UTC on weekdays, plus Chinese holidays) cost half the peak rate. Search and Fetch APIs are a separate meter at $0.75 per 1,000 successful requests.
Tier Strategy
The two models aren't really a good-better-best ladder so much as a speed-versus-capability choice: flash for high-volume, latency-sensitive workloads and v4-pro when output quality matters more than cost per token. There's no seat or plan upgrade to trigger; teams move up when flash starts failing on reasoning-heavy tasks, then often route only those calls to pro. Note that FIM Completion is non-thinking only, so code-completion use cases don't get a reasoning upgrade path.
Competitive Positioning
This is the low-cost challenger play against OpenAI and Anthropic, winning developers and cost-sensitive API shops who care more about price per token than brand safety. The pricing listed here is 'Custom' rather than published, which is a bit surprising for a usage-based API and means you'll have to check their live rate card before benchmarking. The half-price off-peak window is a clear pitch to batch and async workloads that can shift to quieter hours.
Growth Lever
There's no paywall in the traditional sense, since everything is open on metered usage, so expansion is purely volume: more tokens, more Search/Fetch calls, more pro-model traffic. The off-peak discount is the sneaky lever, nudging teams to build scheduling-aware pipelines that increase consumption while feeling like they're saving money. The lack of a free tier or trial means the first dollar of top-up is the real conversion moment, and the granted-balance-first rule softens that for anyone who gets promo credits.
DeepSeek Pricing FAQ
- How much does DeepSeek cost?
- Pricing is custom — you'll need to talk to their sales team for a quote.
- Is there a free plan?
- No free plan. You'll need to commit to a paid plan to get started.
- Can I try it before paying?
- Not at the moment. There's no free trial or free plan listed on their pricing page.
- How does the pricing work?
- Pure pay-as-you-go: you're billed per 1M tokens consumed, with no seats, no subscription floor, and no free tier or trial to ease you in. The twist is time-of-day pricing, where off-peak hours (outside 01:00-04:00 and 06:00-10:00 UTC on weekdays, plus Chinese holidays) cost half the peak rate. Search and Fetch APIs are a separate meter at $0.75 per 1,000 successful requests.
- Is there a discount for annual billing?
- Off-peak rates are half of the peak rates.
- Which plan makes sense for me?
- The two models aren't really a good-better-best ladder so much as a speed-versus-capability choice: flash for high-volume, latency-sensitive workloads and v4-pro when output quality matters more than cost per token. There's no seat or plan upgrade to trigger; teams move up when flash starts failing on reasoning-heavy tasks, then often route only those calls to pro. Note that FIM Completion is non-thinking only, so code-completion use cases don't get a reasoning upgrade path.
More AI pricing
All AI pricing →-
Anthropic Pricing
AI safety company behind Claude.
From $20/mo -
AssemblyAI Pricing
Speech-to-text and audio intelligence API.
See pricing -
Avoma Pricing
AI meeting assistant and revenue intelligence.
From $19/mo Free trial -
Baseten Pricing
ML model deployment infrastructure.
Free plan -
Bland AI Pricing
AI phone agents for automating inbound and outbound calls, priced per minute.
From $0/mo -
Braintrust Pricing
Evals and observability for LLM apps.
From $249/mo -
Claude Pricing
Anthropic's AI assistant with Free, Pro, Max, Team, and Enterprise plans
From $20/mo -
Cohere Pricing
Enterprise AI models and RAG.
See pricing
Weighing these against each other? Track DeepSeek and its alternatives → Pricing, features and reviews, in one weekly brief. Free for 14 days.
Comparing DeepSeek to something specific? Try DeepSeek vs Anthropic, DeepSeek vs AssemblyAI, or DeepSeek vs Avoma. Or see the best DeepSeek alternatives, ranked.
Set it up once. Get a plan every week.
Add the competitors you care about. Outmano keeps watching them, so you can stop checking their sites.
14-day free trial · $0 today