DeepSeek Pricing
All plans and pricing as of 2026-09-14
Free
- ✓Web chat access at chat.deepseek.com
- ✓V4-Flash was the default from the 2026-04-24 launch; V4-Flash retired 2026-09-10 in favour of V4.1-Flash
- ✓Basic usage limits
API -- V4.1-Flash (model name `deepseek-flash`, since 2026-09-10)
- ✓552B total MoE, Causal Encoder-Decoder: 8B active for input, 16B active for output
- ✓1M context, 384K max output, native vision input (no separate Vision model any more)
- ✓PRICE CUT vs the retired V4-Flash: input down ~32% ($0.22 -> $0.15 off-peak), output down ~9% ($0.66 -> $0.60), cache-hit input down ~57% ($0.007 -> $0.003)
- ✓Still above the pre-August flat V4-Flash rate on output ($0.28 -> $0.60 off-peak) -- the 8/16 rise is only partly unwound
- ✓Peak hours 01:00-04:00 and 06:00-10:00 UTC Mon-Fri; all other hours off-peak
- ✓Cache-hit input $0.003 off-peak / $0.006 peak
- ✓Legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp still accepted, served by V4.1-Flash at Flash prices
- ✓Concurrency limit 2500
- ✓Pay-as-you-go, no minimum
API -- V4-Pro
- ✓1.6T total / 49B active MoE
- ✓1M native context
- ✓GA as DeepSeek-V4-Pro-0813 on 2026-08-13
- ✓STILL SERVED AFTER 2026-09-14: DeepSeek announced on 9/10 that all deepseek-v4-pro requests would route to V4.1-Flash from 9/14, then reversed on the change log and rate card -- 'we have decided to continue providing API services for DeepSeek V4 Pro after September 14, 2026, with the billing method remaining unchanged'
- ✓No vision input (V4.1-Flash has it; V4-Pro does not)
- ✓Trails only Gemini 3.1 Pro on world knowledge benchmarks
- ✓PRICING CHANGED 2026-08-16 16:00 UTC (EXECUTED, verified live 8/17): the $0.435/$0.87 standing rate is gone, replaced by peak/off-peak billing at $0.66/$1.98 off-peak and $1.32/$3.96 peak -- a rise at every hour, not a discount. Peak hours are 01:00-04:00 and 06:00-10:00 UTC. See knownIssues
- ✓Still cheaper than Western frontier models, but the gap narrowed sharply on 8/16
Self-hosted (open-source)
- ✓MIT license, open weights on HuggingFace
- ✓V4-Flash is feasible on consumer hardware with quantization
- ✓V4-Pro needs multi-GPU production infrastructure
Is DeepSeek Worth the Price?
Value Score: 9.5/10
Overall Score: 8.0/10 · Developers and teams who need strong reasoning and coding capabilities on a budget. If you're building AI features and can't justify GPT-4 API costs, DeepSeek is the obvious first stop.
DeepSeek is the real deal when it comes to bang-for-your-buck AI. The reasoning capabilities are legitimately impressive, and the open-source angle gives it a flexibility that closed models can't match. The censorship limitations are a dealbreaker for some use cases, and the writing quality trails behind Claude and GPT-4. But for coding, math, and analytical tasks? It's hard to argue with near-frontier performance at a fraction of the cost.
The Tier List Tuesday
Weekly newsletter: tier movers, new entrants, and the VS of the week. Built from our daily AI-tool sweeps. No spam, unsubscribe anytime.
How DeepSeek Pricing Compares
| Tool | Free Tier | Starting Price | Value Score | Overall |
|---|---|---|---|---|
| DeepSeek(this tool) | Yes | $0 | 9.5/10 | 8.0 |
| Qwen (Alibaba) | Yes | $0 | 10/10 | 8.8 |
| MiniMax M3 | Yes | $0 | 9.5/10 | 8.4 |
| Gemma 4 (Google) | Yes | $0 | 10/10 | 8.3 |
| IBM Granite 4.0 | Yes | $0 | 9.5/10 | 8.2 |
| Kimi K3 (Moonshot) | Yes | $3 / $15/per 1M tokens (input/output) | 8.5/10 | 8.1 |