LLM Pricing
ModelsLeaderboardsToolsProvidersReleasesArticles

© 2026 LLM Pricing

About
·Contact
·Privacy
·RSS
← Model list

DeepSeek V4 Pro

deepseek·deepseek/deepseek-v4-pro·GA·Open weights·deepseek-thinking series
Open MoE flagship with million-token context for coding and long agent runs

Specs & pricing

Input / output per 1M tokens
Official price
—
Lowest paid·routing.runGateway
$0.35 / $0.70
Blended $0.44 · 7.4× spread
1 more $0 channels
Context
1,000,000
Output limit
384,000
Knowledge cutoff
2025-05
Released / updated
2026-04-24 / 2026-04-24
Capabilities
✓ Reasoning✓ Tool use✓ Structured output✓ TemperatureAttachments
⚠ Providers report capability flags inconsistently
Modalities
Text
Weights
Hugging Face

DeepSeek V4 Pro is offered by 83 providers on this page. Public prices are shown for 76 of them. No official API price is listed for it; the prices below come from third-party channels. The lowest paid channel is routing.run at $0.348 / $0.696 per 1M. That channel is a third-party gateway, so confirm its availability and rate limits before depending on it. Prices vary about 7.4× across the providers that host it. It also has 1 free ($0) channel and 4 covered by a paid subscription; free tiers usually carry rate limits, and subscription-covered access bills $0 per token only after the subscription fee.

The context window is 1,000,000 tokens at the reference host, but hosts report different limits, from 128,000 to 1,050,000, so the usable window depends on the provider you pick. It supports reasoning, tool use, and structured output. The weights are open, so it can also be self-hosted or served through a gateway of your choice. Providers report the capability flags inconsistently, so verify a specific feature against the host you plan to use. Its training knowledge cuts off at 2025-05.

Available at 83 providers76 with public prices · 1 free · 4 subscription-covered

  • ProviderInput / Output
  • SCNet Token PlanGateway
    DeepSeek-V4-Pro
    Subscription
    Cache read $0Context 1,000,000Output limit 384,000
  • Alibaba Token PlanGateway
    Subscription
    Cache read $0Context 1,000,000Output limit 384,000
  • Alibaba Token Plan (China)Gateway
    Subscription
    Cache read $0Context 1,000,000Output limit 384,000
  • UnoRouterGateway
    deepseek-v4-pro:free
    Free
    Context 1,000,000Output limit 384,000deprecated
  • Volcengine Ark Coding PlanGateway
    Subscription
    Cache read $0Context 1,000,000Output limit 384,000
  • routing.runGateway
    $0.35 / $0.70
    Context 1,000,000Output limit 64,000
  • EBCloudGateway
    DeepSeek-V4-Pro
    $0.43 / $0.86
    Context 1,000,000Output limit 384,000
  • ModelisGateway
    $0.44 / $0.87
    Context 1,000,000Output limit 384,000
  • Hugging FaceCloud
    deepseek-ai/DeepSeek-V4-Pro
    $0.44 / $0.87
    Cache read $0.0036Context 1,048,576 ⚠Output limit 393,216
  • TokenGoGateway
    $0.44 / $0.87
    Context 1,000,000Output limit 384,000
ProviderTierInputOutputCache readCache writeContextOutput limitStatus
SCNet Token Plan
DeepSeek-V4-Pro
GatewaySubscription$0—1,000,000384,000
Alibaba Token PlanGatewaySubscription$0$01,000,000384,000
Alibaba Token Plan (China)GatewaySubscription$0$01,000,000384,000
UnoRouter
deepseek-v4-pro:free
GatewayFree——1,000,000384,000deprecated
Volcengine Ark Coding PlanGatewaySubscription$0—1,000,000384,000
routing.runGateway$0.35$0.70——1,000,00064,000
EBCloud
DeepSeek-V4-Pro
Gateway$0.43$0.86——1,000,000384,000
ModelisGateway$0.44$0.87——1,000,000384,000
Hugging Face
deepseek-ai/DeepSeek-V4-Pro
Cloud$0.44$0.87$0.0036—1,048,576 ⚠393,216
TokenGoGateway$0.44$0.87——1,000,000384,000

Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.

Reasoning control

Toggle (on / off)effort = higheffort = maxeffort = noneeffort = loweffort = minimaleffort = mediumeffort = xhighbudget_tokens ≥ 1 ≤ 393,216

Interleaved thinking (reasoning between tool calls) is declared by 55 of 83 providers.

11 / 83 providers expose no reasoning control (reasoning_options: []).

Your usage cost

Common scenarios
1Hugging Face$78.74
2AIHubMix$86.56
3APIKO$94.77
4Kenari$103.40
5routing.run$104.40
6EBCloud$128.58

5 more channels offer $0 (SCNet Token Plan, Alibaba Token Plan, Alibaba Token Plan (China) etc.); free tiers usually have rate limits and no SLA, excluded from ranking.

The cheapest paid channel is the only channel.

Price historyone sample accumulated per data sync

Input listOutput listMin blended
$1.98$02026-08-052026-10-26
2026-10-26Input list —Output list —Min blended $0.44

Benchmark26 items

NameConditionsScoreMetricSource
SWE-Bench Verified—80.6resolvedSource ↗
Artificial Analysis Coding Agent Indexharness: Claude Code · variant: high50.1average pass@1Source ↗
SWE-Atlas Codebase QnAharness: Claude Code · variant: high67.8pass@1Source ↗
SWE-Bench Proharness: Claude Code · variant: high · dataset: hard-aa18pass@1Source ↗
Terminal-Benchharness: Claude Code · variant: high · v2.164.7pass@1Source ↗
MMLU-Provariant: preview checkpoint; max effort87.5EMSource ↗
SimpleQA-Verifiedvariant: preview checkpoint; max effort57.9pass@1Source ↗
Chinese SimpleQAvariant: preview checkpoint; max effort84.4pass@1Source ↗
GPQA Diamondvariant: preview checkpoint; max effort90.1pass@1Source ↗
Humanity's Last Examvariant: preview checkpoint; max effort; without tools37.7pass@1Source ↗
LiveCodeBenchvariant: preview checkpoint; max effort93.5pass@1Source ↗
Codeforcesvariant: preview checkpoint; max effort3206ratingSource ↗
HMMTvariant: preview checkpoint; max effort · dataset: February 202695.2pass@1Source ↗
IMOAnswerBenchvariant: preview checkpoint; max effort89.8pass@1Source ↗
MathArena Apexvariant: preview checkpoint; max effort38.3pass@1Source ↗
MathArena Apex Shortlistvariant: preview checkpoint; max effort90.2pass@1Source ↗
MRCRvariant: preview checkpoint; max effort · dataset: 1M context83.5MMRSource ↗
CorpusQAvariant: preview checkpoint; max effort · dataset: 1M context62accuracySource ↗
Terminal-Benchvariant: preview checkpoint; max effort · v2.067.9accuracySource ↗
SWE-Bench Provariant: preview checkpoint; max effort55.4resolvedSource ↗
SWE-Bench Multilingualvariant: preview checkpoint; max effort76.2resolvedSource ↗
BrowseCompvariant: preview checkpoint; max effort83.4pass@1Source ↗
Humanity's Last Examvariant: preview checkpoint; max effort; with tools48.2pass@1Source ↗
MCP Atlasvariant: preview checkpoint; max effort · dataset: public73.6pass@1Source ↗
GDPval-AAvariant: preview checkpoint; max effort1554EloSource ↗
Toolathlonvariant: preview checkpoint; max effort51.8pass@1Source ↗

The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.

Related models

DeepSeek V4 Pro 0813same series$0.66 / $1.98DeepSeek V4 Pro 0423same series—DeepSeek Reasonersame series—DeepSeek V4 Flash Vision Expsame lab—
Data partly from models.dev (MIT)