LLM Reference

DeepSeek V4 Flash vs DeepSeek V4 Pro

DeepSeek V4 Flash (2026) and DeepSeek V4 Pro (2026) are frontier-tier reasoning models from DeepSeek. DeepSeek V4 Flash ships a 1m-token context window, while DeepSeek V4 Pro ships a 1m-token context window. On MMLU PRO, DeepSeek V4 Pro leads by 1.1 pts. This comparison covers specs, pricing, API access, capabilities, benchmarks, input and output token costs, and production fit for coding and agent workloads.

DeepSeek V4 Flash is safer overall; choose DeepSeek V4 Pro when provider fit matters.

Decision scorecard

Local evidence first
SignalDeepSeek V4 FlashDeepSeek V4 Pro
Best forreasoning-heavy apps, tool-calling agents, and long-context analysisreasoning-heavy apps, tool-calling agents, and long-context analysis
Decision fitCoding, RAG, and AgentsCoding, RAG, and Agents
Context window1m1m
Cheapest output$0.12/1M tokens$0.87/1M tokens
Provider routes5 tracked5 tracked
Shared benchmarks15 sharedMMLU PRO leader

Decision tradeoffs

Choose DeepSeek V4 Flash when...
  • DeepSeek V4 Flash has the lower cheapest tracked output price at $0.12/1M tokens.
  • Local decision data tags DeepSeek V4 Flash for Coding, RAG, and Agents.
Choose DeepSeek V4 Pro when...
  • DeepSeek V4 Pro holds a shared-benchmark lead on MMLU PRO, ahead by 1.1 points.
  • Local decision data tags DeepSeek V4 Pro for Coding, RAG, and Agents.

Monthly cost at traffic

Estimate token spend from the cheapest tracked input and output route or tier on this page.

Lower estimate DeepSeek V4 Flash

DeepSeek V4 Flash

$76.26

Cheapest tracked route/tier: OpenRouter

DeepSeek V4 Pro

$566

Cheapest tracked route/tier: DeepSeek Platform

Estimated monthly gap: $489. Batch, cache, alternate speed tiers, and negotiated pricing are excluded from this local estimate.

Switch friction

DeepSeek V4 Flash -> DeepSeek V4 Pro
  • Provider overlap exists on DeepSeek Platform, OpenRouter, and Vercel AI Gateway; start route-level A/B tests there.
  • DeepSeek V4 Pro is $0.75/1M tokens higher on cheapest tracked output pricing, so quality gains need to justify the spend.
DeepSeek V4 Pro -> DeepSeek V4 Flash
  • Provider overlap exists on DeepSeek Platform, OpenRouter, and Vercel AI Gateway; start route-level A/B tests there.
  • DeepSeek V4 Flash is $0.75/1M tokens lower on cheapest tracked output pricing before cache, batch, or negotiated discounts.

Specs

Specification
Released2026-04-242026-04-24
Context window1m1m
Parameters284B1.6T
ArchitectureMixture of ExpertsMixture of Experts
LicenseMITOSI-approvedMITOSI-approved
OpennessOpen sourceOpen source
WeightsAvailableAvailable
CodeUnknownUnknown
Commercial useCommercial use: permittedCommercial use: permitted
Knowledge cutoff--

Pricing and availability

Pricing attributeDeepSeek V4 FlashDeepSeek V4 Pro
Input price
Off-peak
$0.22/1M tokens
All UTC hours outside DeepSeek's peak windows. Cache-hit input: $0.007 per 1M tokens; cache-miss input: $0.22 per 1M; output: $0.66 per 1M.
Peak
$0.44/1M tokens
01:00-04:00 and 06:00-10:00 UTC. Cache-hit input: $0.014 per 1M tokens; cache-miss input: $0.44 per 1M; output: $1.32 per 1M.
$0.43/1M tokens
Output price
Off-peak
$0.66/1M tokens
All UTC hours outside DeepSeek's peak windows. Cache-hit input: $0.007 per 1M tokens; cache-miss input: $0.22 per 1M; output: $0.66 per 1M.
Peak
$1.32/1M tokens
01:00-04:00 and 06:00-10:00 UTC. Cache-hit input: $0.014 per 1M tokens; cache-miss input: $0.44 per 1M; output: $1.32 per 1M.
$0.87/1M tokens
Providers
DeepSeek PlatformOpenRouterMicrosoft FoundryVercel AI GatewayNovita AI
DeepSeek PlatformFireworks AIOpenRouterVercel AI GatewayNovita AI

Capabilities

CapabilityDeepSeek V4 FlashDeepSeek V4 Pro
VisionNoNo
MultimodalNoNo
ReasoningYesYes
JSON / Tool useYesYes
Structured outputsYesYes
Code executionNoNo
IDE integrationNoNo
Computer useNoNo
Parallel agentsNoNo

Benchmarks

BenchmarkDeepSeek V4 FlashDeepSeek V4 Pro
MMLU PRO86.487.5
SWE-bench Verified79.080.6
SWE-bench Pro52.655.4
Google-Proof Q&A88.190.1
LiveCodeBench91.693.5
Humanity's Last Exam34.837.7
HumanEval69.576.8
Chatbot Arena1437.01456.0
BrowseComp73.283.4
Mathematics Aptitude Test of Heuristics57.464.5
SWE-bench Multilingual73.376.2
Terminal-Bench 2.056.959.1
Massive Multitask Language Understanding88.790.1
MCP-Atlas69.069.4
GeneBench-Pro2.42.4

Continue comparing

Model pages
Labs and families
Related comparisons

Popular comparisons for DeepSeek V4 Flash

Popular comparisons for DeepSeek V4 Pro

Last reviewed: 2026-07-03. Data sourced from public model cards and provider documentation.