DeepSeek V4 Pro pricing & benchmarks
DeepSeek V4 Pro is available from DeepSeek at $0.435 per million input tokens and $0.870 per million output tokens ($0.544 blended at 3:1). It accepts up to 1,048,576 tokens of context. DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...
Pricing
Per million tokens · OpenRouter
Input$0.435
Output$0.870
Blended (3:1)$0.544
Specification
deepseek/deepseek-v4-pro
Context window1.0M
Max output384K
ReleasedApr 2026
LicenceOpen weights
Modalitiesin:text, out:text
Benchmarks
Independently run — not vendor self-reports. Every score links its source.
| Benchmark | Score | Config | Run | Source |
|---|---|---|---|---|
| GPQA Diamond | 90.9% ±2.0 | high effort | 2026-08-06 | Independently run · Epoch AI |
| SWE-bench Verified | 77.6% ±1.9 | max effort | 2026-06-18 | Independently run · Epoch AI |
| FrontierMath T1–3 | 45.3% ±3.0 | max effort | 2026-06-17 | Independently run · Epoch AI |
| FrontierMath T4 | 2.4% ±2.4 | max effort | 2026-06-17 | Independently run · Epoch AI |
| SimpleQA Verified | 57.0% ±1.6 | max effort | 2026-06-16 | Independently run · Epoch AI |
| OTIS Mock AIME | 96.7% ±2.0 | max effort | 2026-06-17 | Independently run · Epoch AI |
Scores published by Epoch AI, 'AI Benchmarking Hub', used under CC BY 4.0.
Cheaper models that score at least as well on GPQA Diamond
Strictly lower blended price AND an equal-or-higher GPQA Diamond score. One benchmark is one dimension — a model that wins here may still be weaker at your specific task.
| Model | Provider | Blended $/1M | GPQA Diamond | |
|---|---|---|---|---|
| DeepSeek V4 Flash 0731 | DeepSeek | $0.113 (−79%) | 91.0% | Compare → |
| GPT-5.6 Luna | OpenAI | $0.225 (−59%) | 91.6% | Compare → |
| GLM 5.2 | Zhipu AI | $0.387 (−29%) | 91.9% | Compare → |
Where to run DeepSeek V4 Pro
18 hosts · 5.6× between cheapest and dearest. Blended at 3:1. ⚠ marks an option serving less context than the best available, or re-pricing above a prompt-length threshold. Bracketed labels are the provider's own service tier or region.
| Host | Blended $/1M | In / Out | Context | Weights | Uptime 24h |
|---|---|---|---|---|---|
| StreamLake | $0.390 | $0.312 / $0.625 | 1.0M | fp8 | 97.2% |
| Baidu | $0.401 | $0.321 / $0.642 | 1.0M | fp8 | 99.6% |
| DeepSeek | $0.544 | $0.435 / $0.870 | 1.0M | — | 99.9% |
| Novita | $0.749 | $0.600 / $1.20 | 1.0M | fp8 | 99.8% |
| GMICloud | $0.827 | $0.661 / $1.32 | 1.0M | fp8 | 98.9% |
| DigitalOcean | $1.09 | $0.870 / $1.74 | 1.0M | — | 98.1% |
| Ionstream | $1.41 | $1.13 / $2.26 | 1.0M | fp4 | 97.9% |
| Cloudflare ⚠ 393K context, not 1.0M |
$1.46 | $1.17 / $2.34 | 393K | — | 99.0% |
| CoreWeave | $1.50 | $1.15 / $2.55 | 1.0M | fp8 | 99.6% |
| DeepInfra | $1.63 | $1.30 / $2.60 | 1.0M | fp4 | 99.0% |
| Alibaba | $1.77 | $1.42 / $2.83 | 1M | fp8 | 99.5% |
| SiliconFlow | $1.91 | $1.50 / $3.14 | 1.0M | fp8 | 98.3% |
| Venice | $2.06 | $1.65 / $3.30 | 1M | — | 65.9% |
| AtlasCloud | $2.10 | $1.68 / $3.38 | 1.0M | fp4 | 99.3% |
| BaseTen ⚠ 262K context, not 1.0M |
$2.17 | $1.74 / $3.48 | 262K | fp4 | 99.8% |
| Parasail | $2.17 | $1.74 / $3.48 | 1.0M | fp8 | 95.2% |
| Together ⚠ 512K context, not 1.0M |
$2.17 | $1.74 / $3.48 | 512K | — | 91.6% |
| Fireworks | $2.17 | $1.74 / $3.48 | 1.0M | — | 0.0% |
Host prices and uptime from OpenRouter, synced 2026-08-08. Lower-precision weights (fp4, fp8) can cost less and score worse than the same model at bf16 — the score above is not host-specific.
Subscriptions that include DeepSeek V4 Pro
Consumer plans whose published model list names this model
Compare DeepSeek V4 Pro
FAQ
How much does DeepSeek V4 Pro cost?
DeepSeek V4 Pro costs $0.435 per million input tokens and $0.870 per million output tokens on OpenRouter. At a 3:1 input:output mix that blends to $0.544 per million tokens.
What is DeepSeek V4 Pro's context window?
DeepSeek V4 Pro accepts up to 1,048,576 tokens of context, and can generate up to 384,000 output tokens.
How good is DeepSeek V4 Pro on benchmarks?
DeepSeek V4 Pro scores 90.9% on GPQA Diamond, independently run and published by Epoch AI. 6 benchmark results are listed on this page, each with its run date and source.
Is there a cheaper model as good as DeepSeek V4 Pro?
Yes — 3 models in our catalogue cost less per token than DeepSeek V4 Pro and score at least as high on the same benchmark. The cheapest is DeepSeek V4 Flash 0731 at $0.113 per million tokens blended.