DeepSeek V3 API pricing
DeepSeek · Budget price tier · DeepSeek License · released 2024-12. Prices in USD per 1 million tokens. Data snapshot: 2026-11-14.
Promotional rate. The price we quote is 10% off StreamLake's standard $0.286 per 1M input tokens and $1.143 per 1M output tokens, and a promotion can end without notice.
A support ticket workload of 100K requests a month costs $90.05 on DeepSeek V3; the next cheaper model in the Budget price tier for this workload, Codestral 2508, does it for $90.00. The price we quote for DeepSeek V3 has changed 8 times since 2026-08-15, most recently on 2026-11-14 when the quoted host changed (+10.9%); it is now the 119th cheapest of 159 models in the Budget price tier. Buying from DeepInfra rather than at the price we quote lowers the token price by about 10%, for a quantized fp4 build. It is rated a stable release but does not meet all four criteria of the FinOps Friendly badge.
- Input: $0.2574 per 1M tokens
- Output: $1.0287 per 1M tokens
- Batch API (in/out): Not published
- Prompt-cache read: Not published
- Context window: 164K tokens
- Arena ELO: 1333 as of 2026-11-09
Market position
On blended price ($0.7973 per 1M tokens at 30% input, 70% output) it is cheaper than 25% of models in the Budget price tier: 119th cheapest of 159. Across every price tier it is cheaper than 61% of priced text models in the catalogue: 119th cheapest of 304. Its Arena ELO of 1333 ranks 129th of the 160 models that carry one, tied with one other, above 19% of them. With an Arena score of 1333, DeepSeek V3 sits below OptimNow's lowest working quality threshold, Standard (Arena 1420+), so the quality-bar method would not shortlist it for any of the 8 use cases the table prices. Each threshold is a distance below the best ranked model, which scores 1530 on Arena scores as of 9 Nov 2026.
Licence, size and capabilities
DeepSeek V3 is open weights, which is not the same as open source: the weights are published, but under DeepSeek License, a licence that restricts some uses. Read it before self-hosting commercially. Its parameter count is not published. Listed capabilities: Text, Agents. Filling its 164K context window once costs $0.0422 in input tokens at list price.
Pricing by host
A host is a company that runs the model and sells it by the token: the model maker, a cloud platform like Amazon Bedrock, Azure or Vertex, or an independent host. One model, several hosts, and often several prices. This model is sold by 2 different hosts. The price we quote elsewhere on this page is marked. We quote one host's rate so that models stay comparable; it is not a recommendation to buy there.
Cheapest host: DeepInfra at $0.32 input / $0.89 output per 1M tokens, 10% below the price we quote on the blended price (30% input, 70% output). It serves a quantized fp4 build, which is not the same product as the full-precision model. The dearest, StreamLake, charges 1.1x the cheapest.
| Host | Input $/1M | Output $/1M | Precision |
|---|---|---|---|
| DeepInfra | $0.32 | $0.89 | fp4 |
| StreamLake (price we quote) | $0.2574 | $1.0287 | not declared |
List prices as published on 2026-11-14, via OpenRouter. Negotiated rates, committed-use discounts and provisioned capacity are not shown. StreamLake's listing is a promotional rate: 10% off its standard rate.
Monthly cost at 100K requests, by business use case
From the after-discounts figure (prompt caching on the cached input share, batch API for async workloads, only where DeepSeek publishes those rates), which assumes you implement both, to the list price. One figure means no discount is published.
| Use case | Tokens (in/out) | Monthly (after discounts – list) |
|---|---|---|
| Support Ticket | 1,500 / 500 | $90 |
| Knowledge Q&A | 2,000 / 800 | $134 |
| Meeting Summary | 10,000 / 1,200 | $381 |
| Marketing Content | 2,500 / 1,800 | $250 |
| Coding Task | 3,000 / 2,000 | $283 |
| Invoice Processing | 1,500 / 600 | $100 |
| Call Summary | 2,000 / 700 | $123 |
| Agent Workflow | 6,000 / 3,000 | $463 |
Price history
DeepSeek V3 has been in our daily price archive since tracking began on 2026-08-15, when it listed at $0.2574 input / $1.0287 output per 1M tokens. The price we quote has changed 8 times since then, every time through a change of quoted host rather than a repricing. The latest change, on 2026-11-14, was +10.9% on the blended price (30% input, 70% output), to $0.2574 input / $1.0287 output (quoted host changed, not a repricing). Net of all changes, the blended price is back where it started on 2026-08-15.
| Date | Old price (in / out) | New price (in / out) | Blended change |
|---|---|---|---|
| 2026-11-14 | $0.32 / $0.89 | $0.2574 / $1.0287 | +10.9% (quoted host changed, not a repricing) |
| 2026-11-06 | $0.2574 / $1.0287 | $0.32 / $0.89 | -9.8% (quoted host changed, not a repricing) |
| 2026-10-20 | $0.32 / $0.89 | $0.2574 / $1.0287 | +10.9% (quoted host changed, not a repricing) |
| 2026-10-09 | $0.2574 / $1.0287 | $0.32 / $0.89 | -9.8% (quoted host changed, not a repricing) |
| 2026-09-28 | $0.32 / $0.89 | $0.2574 / $1.0287 | +10.9% (quoted host changed, not a repricing) |
3 earlier changes are not listed.
Alternatives to DeepSeek V3
3 of the 4 comparable models below cost less than DeepSeek V3: Gemma 3 12B undercuts it by 85%, and scores 1 Arena point higher.
- Next cheaper model in the Budget price tier: Qwen3 Next 80B A3B Instruct (Alibaba) at $0.797 blended, under 1% less.
- Next more expensive: Llama 4 Maverick (Meta) at $0.805 blended, 1% more.
The nearest models in the Budget price tier on Arena ELO, each with its blended price and score difference against DeepSeek V3.
- GLM 4.5V (Zhipu): $0.60 in / $1.80 out per 1M, Arena ELO 1333 (81% more expensive, same Arena ELO)
- Gemma 3 12B (Google): $0.05 in / $0.15 out per 1M, Arena ELO 1334 (85% cheaper, 1 ELO point higher)
- Nemotron 3.5 Lightning (Nvidia): $0.07 in / $0.20 out per 1M, Arena ELO 1329 (80% cheaper, 4 ELO points lower)
- Mistral Small 3.2 24B (Mistral): $0.10 in / $0.30 out per 1M, Arena ELO 1338 (70% cheaper, 5 ELO points higher)
- Cheapest within 15 Arena points: Gemma 3 12B (Google), 85% cheaper, 1 ELO point higher, out of 11 models in that band.
- Highest Arena ELO within 20% of this price: MiMo-V2.6-Pro (Xiaomi), 7% cheaper, 159 ELO points higher, out of 23 models in that price band.
Other models from DeepSeek
The 8 nearest to DeepSeek V3 in price, of 13 other DeepSeek models in the catalogue. Prices per 1M tokens, input then output.
- DeepSeek V4.1 Flash 1017: $0.038 / $0.50
- DeepSeek V3.2 Exp: $0.27 / $0.41
- DeepSeek V3.1: $0.25 / $0.95
- DeepSeek V3.1 Terminus: $0.27 / $1.00
- DeepSeek V3 0324: $0.29 / $1.14
- DeepSeek V4 Flash Vision Exp: $0.44 / $1.32
- DeepSeek V4 Pro 0423: $0.90 / $1.50
- R1 0528: $0.50 / $2.15
All 14 DeepSeek models and prices
Common questions about DeepSeek V3 pricing
- How much does DeepSeek V3 cost?
- DeepSeek V3 costs $0.2574 per 1M input tokens and $1.0287 per 1M output tokens at StreamLake, the host whose rate we quote. Prices as listed on 2026-11-14.
- Which host is cheapest for DeepSeek V3?
- Cheapest host: DeepInfra at $0.32 input / $0.89 output per 1M tokens, 10% below the price we quote on the blended price (30% input, 70% output). It serves a quantized fp4 build, which is not the same product as the full-precision model. The dearest, StreamLake, charges 1.1x the cheapest. Host rates as listed on 2026-11-14.
- Does DeepSeek V3 offer batch or cached pricing?
- No. Neither a batch rate nor a prompt-cache rate is published for DeepSeek V3 as of 2026-11-14.
- What would DeepSeek V3 cost per month?
- With the Support Ticket profile (1,500 input and 500 output tokens per request), 100,000 requests a month cost about $90 on DeepSeek V3 at the list price quoted on 2026-11-14.
- Can DeepSeek V3 be self-hosted?
- Yes, within the terms of its licence. DeepSeek V3 is open weights, which is not the same as open source: the weights are published, but under DeepSeek License, a licence that restricts some uses. Read it before self-hosting commercially.
- Has the price of DeepSeek V3 changed?
- DeepSeek V3 has been in our daily price archive since tracking began on 2026-08-15, when it listed at $0.2574 input / $1.0287 output per 1M tokens. The price we quote has changed 8 times since then, every time through a change of quoted host rather than a repricing. The latest change, on 2026-11-14, was +10.9% on the blended price (30% input, 70% output), to $0.2574 input / $1.0287 output (quoted host changed, not a repricing). Net of all changes, the blended price is back where it started on 2026-08-15.
Its weights are published, which changes what you may run yourself: the cost gap between open weights and proprietary APIs.
These are list prices. Your real cost also depends on caching and batching, negotiated discounts, marketplace agreements, and the contracts you already have. OptimNow audits exactly that.
Build the business case · Get a FinOps review
LLM pricing · Models by maker · Cloud compute pricing · Compare models · Price barometer · FinOps guides · API and MCP server