Skip to main content

DeepSeek V4 Pro

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, and long-horizon agent workflows, with strong performance across knowledge, math, and software engineering benchmarks. Built on the same architecture as DeepSeek V4 Flash, it introduces a hybrid attention system for efficient long-context processing. Reasoning efforts high and xhigh are supported; xhigh maps to max reasoning. It is well suited for complex workloads such as full-codebase analysis, multi-step automation, and large-scale information synthesis, where both capability and efficiency are critical.

Model card on Hugging Face
Input / output price
$0.90 / $1.50 per 1M
Context
1M tokens
Released
Apr 24, 2026
Agent setup

Paste into Claude Code, Cursor, or Codex

Help me set up the Respan gateway: fetch and follow https://www.respan.ai/docs/agent-setup.md. Use model "deepseek/deepseek-v4-pro".

Read docs

Providers

16 providers, cheapest first, retired then OpenRouter last, priced per 1M tokens after gateway discounts.

ProviderInputOutput
Waferwafer/deepseek-v4-pro
$0.28/M$3.96/M
Relacerelace/deepseek-v4-pro
$0.45/M$3.00/M
Concentrateconcentrate/deepseek-v4-pro
$0.50/M$2.00/M
Deepseekdeepseek/deepseek-v4-pro
$0.66/M$1.98/M
DeepInfradeepinfra/deepseek-v4-pro
$1.30/M$2.60/M

Is a provider down? DeepSeek status, OpenRouter status

Pricing

Averaged across 16 providers, in USD per 1M tokens after gateway discounts.

Average input
$1.32 per 1M
Average output
$3.07 per 1M
Average cache read
$0.30 per 1M

Cost by volume

Total cost from 0 to 1T tokens on the cheapest and most expensive provider.

Highest$2.10/MInference.netLowest$0.28/MWafer

Throughput and latency

30-day averages from requests through the Respan gateway. Under each figure, every model on the gateway by its average, this one in blue.

Output speed

Higher is better

104tokens/sacross providers

Faster than 79% of the other 596 models on the gateway

Time to first token

Lower is better

1.8sacross providers

Faster than 46% of the other 153 models on the gateway

Response latency

Lower is better

6.2sacross providers

Faster than 27% of the other 768 models on the gateway

Benchmarks

The Respan Index plus 31 published scores from 13 sources, each linked to where it was published.

49.2/ 100

Respan Index#45 of 176

176 ranked models. The shaded band is this model's 95% interval, 43.8 to 54.9. GPT-5.4 and GPT-6 Luna fall inside it, so the gap is within the margin of error.
DeepSeek V4 Pro by category: Coding 29.8, Agents 59.7, Reasoning 52.8, Knowledge & language 52.6, Human preference 55.7
  1. 45DeepSeek V4 Pro49.2
  2. 44GPT-5.449.9
  3. 46GPT-6 Luna48.6

Published scores

BenchmarkScore
Terminal-Bench 2.1DeepSeek87.9%
SWE-bench VerifiedEpoch AI77.6%
LiveBench CodingLiveBench77.2%
WeirdMLHåvard Tveit Ihle66.2%
DeepSWE v1.1Datacurve62.8%
DeepSWE v1.1DeepSeek62.7%
SWE-rebench 2026-05-15 to 2026-07-01Nebius40.2%
MirrorCodeEpoch AI33.3%
Terminal-Bench 4.0Vals AI14.1%

Compare DeepSeek V4 Pro: DeepSeek V4 Pro vs Claude Sonnet 5.5, DeepSeek V4 Pro vs Gemini 3.8 Flash, DeepSeek V4 Pro vs DeepSeek-V4.1-Flash-1017, DeepSeek V4 Pro vs DeepSeek V4.1 Flash, DeepSeek V4 Pro vs DeepSeek V4 Flash, DeepSeek V4 Pro vs DeepSeek-V4-Flash (Preview)

Apps

Use DeepSeek V4 Pro in the tools you already run. Each guide points the app at the Respan gateway.

Setup guides for more apps are in the integration docs.

DeepSeek V4 Pro FAQ

Short answers, built from the data on this page.

DeepSeek V4 Pro is a model from Deepseek, released on Apr 24, 2026. 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, and long-horizon agent workflows, with strong performance across knowledge, math, and software engineering benchmarks. Built on the same architecture as DeepSeek V4 Flash, it introduces a hybrid attention system for efficient long-context processing. Reasoning efforts high and xhigh are supported; xhigh maps to max reasoning. It is well suited for complex workloads such as full-codebase analysis, multi-step automation, and large-scale information synthesis, where both capability and efficiency are critical.

On the Respan gateway, input starts at $0.28 per 1M tokens on Wafer, and output starts at $1.50 per 1M tokens on OpenRouter, after gateway discounts. Averaged across its 16 providers, it costs $1.32 per 1M input tokens and $3.07 per 1M output tokens.

DeepSeek V4 Pro is available through 16 providers: Wafer, Relace, Concentrate, Deepseek, DeepInfra, IO.net, and 10 more.

DeepSeek V4 Pro accepts up to 1M tokens of context on its largest provider route.

Point any OpenAI-compatible SDK at https://gateway.respan.ai/v1 with your Respan API key and set the model to "deepseek/deepseek-v4-pro". To set it up with a coding agent, copy the agent setup prompt at the top of this page.

DeepSeek V4 Pro scores 49.2 out of 100 on the Respan Index v1.1, ranking #45 of 176 models. By category: Coding 29.8, Agents 59.7, Reasoning 52.8, Knowledge & language 52.6, and Human preference 55.7. The Respan Index combines independent and lab-reported benchmark results into one score, using benchmark data collected by Respan.

DeepSeek V4 Pro has 31 published benchmark scores, including Terminal-Bench 2.1 87.9%, Toolathlon-Verified 74.4%, and GPQA Diamond 91.7%. Each score links to where it was published; the data is collected by Respan.

Certified where it counts.

Built to meet the security and privacy standards that enterprise and healthcare teams require.

ISO 27001

The internationally recognized standard for information security management.

SOC 2

Secure, compliant management of your data across all of our systems.

GDPR

Operated under GDPR, the world's strictest standard for data privacy.

HIPAA

HIPAA compliant, with a BAA available for healthcare teams.

Works with your entire stack

Use Respan with your favorite frameworks and tools.

AnthropicAI21 LabsxAIVercel AI SDK
LangChainOpenRouterAssemblyAICursor
LiteLLMGroqBasetenMem0
CohereCogneeLlamaIndexFireworks
DeepSeekLinkupHaystackTogether AI
PerplexityInference.netPostHogRubyLLM
SlackMastraAzure OpenAIMistral
AWS BedrockMoonshotHyperspellBAML
ModaInstructorGoogle GeminiParasail
ReductoZapierAgnoNebius
OpenAINovita AIReplicateSuperagent

Built for AI agents.
Break less.
Ship more.