Official V4-Flash-0731 release July 31, 2026; weights on Hugging Face, API in public beta. Only DeepSeek model supporting the Responses API. DeepSeek has warned of a future price increase.
Terminal-Bench 2.1 82.7 — up from 61.8 in the April preview, beating V4-Pro (Preview) on all nine published agent benchmarks
Strong tool-calling and security-task results (Toolathlon-Verified 70.3, Cybergym 76.7)
$0.14/$0.28 per 1M with 1M context and MIT-licensed weights
Weaknesses
Well behind GPT-5.6, Opus-class, and Gemini frontier models on the hardest reasoning and long-horizon work
Text-only, and several headline numbers come from DeepSeek's own unreleased eval framework
Real-world use cases
What people actually use DeepSeek V4-Flash for.
Agent pipelines at $0.14/1M input — Terminal-Bench 2.1 82.7 rivals models 30x its price
Tool-calling workloads (Toolathlon-Verified 70.3) with 2,500 concurrent requests
Self-hosting in ~110 GB at 3-bit quantization under MIT license
How DeepSeek V4-Flash compares
The nearest models people weigh against it, and what actually separates them.
vs DeepSeek V4-Pro — Against DeepSeek V4-Pro (DeepSeek), DeepSeek V4-Flash runs about 68% cheaper per token and answers faster. Take DeepSeek V4-Flash unless you specifically need what DeepSeek V4-Pro does better.
vs Devstral Small 1.1 — Against Devstral Small 1.1 (Mistral), DeepSeek V4-Flash costs about 5% more per token and takes 7.6x the context. Devstral Small 1.1 is the one to check first if the price difference matters more than the ceiling.
vs GLM-5.2 — Against GLM-5.2 (Z.ai), DeepSeek V4-Flash runs about 93% cheaper per token and answers faster. Take DeepSeek V4-Flash unless you specifically need what GLM-5.2 does better.
Price History
DeepSeek V4-Flash pricing over time
→0% since Aug 7
38 data points · tracked daily since Aug 7, 2026
Ready to try it?
Start using DeepSeek V4-Flash
High-volume agentic coding and tool-use pipelines. Start free — no card required.
DeepSeek's 1.6T-parameter (49B active) MoE flagship with hybrid sparse attention — near-frontier coding and reasoning at roughly a tenth of closed-rival pricing, MIT-licensed open weights.
Verdict
Best open-weights flagship — near-frontier coding at a tenth of the price.
Quality score
82%
Pricing
$0.43/1M in
$0.87/1M out
Speed
Balanced
3/5 speed
Context
1M tokens
Open-weight preview April 24; GA ~July 20, 2026. Off-peak pricing verified on api-docs.deepseek.com; Beijing-business-hours surge doubles it. Legacy deepseek-chat/reasoner endpoints retired July 24, 2026.
Devstral Small 1.1 is Mistral's code-specialized small model, purpose-built for software engineering tasks including code generation, debugging, and repository-level reasoning. It succeeds Devstral Small 1.0 with improved instruction following and agentic coding capabilities at a fraction of flagship model costs.
Verdict
The best dollar-for-dollar coding model for agentic pipelines that doesn't need to do anything else.
Quality score
54%
Pricing
$0.10/1M in
$0.30/1M out
Speed
Fast
4/5 speed
Context
131k tokens
Available via Mistral API and can be self-hosted via open weights. Pricing is among the lowest available for a code-specialized model. Designed to work within coding agent frameworks like SWE-agent and OpenHands.
Z.ai's MIT-licensed open-weight flagship — the top open-weights coding model of mid-2026, beating GPT-5.5 on agentic coding benchmarks at roughly a sixth of the cost.
Verdict
Top open-weights coder — beats GPT-5.5 at a sixth of the cost.
Quality score
80%
Pricing
$1.40/1M in
$4.40/1M out
Speed
Balanced
3/5 speed
Context
1M tokens
Announced June 13, 2026; pay-per-token API live June 16. Two reasoning modes ('thinking' and 'max thinking'). GLM Coding Plan: Lite $18/mo (~$12.60 effective yearly), Pro $72, Max $160.
DeepSeek V4-Flash costs $0.14 per million input tokens and $0.28 per million output tokens on the API. A month of 10M input and 2M output tokens runs about $1.96 at list price, before any batch or caching discounts.
What is the context window of DeepSeek V4-Flash?
DeepSeek V4-Flash has a 1M tokens context window, with up to 384k tokens of output per response. That is the total of prompt plus response the model can hold in one request.
What is the knowledge cutoff of DeepSeek V4-Flash?
DeepSeek V4-Flash's training data runs through May 2025, and the model was released on April 23, 2026. For anything after that date it needs web search or documents in the prompt.
What is DeepSeek V4-Flash best for?
DeepSeek V4-Flash is best for high-volume agentic coding and tool-use pipelines. It is a strong fit when that workflow matters more than the tradeoffs around budget pricing and fast speed.
When should I avoid DeepSeek V4-Flash?
You need vision input or frontier-grade reasoning on the hardest tasks.
What is a cheaper alternative to DeepSeek V4-Flash?
Devstral Small 1.1 (Mistral) at $0.10/1M/1M input against DeepSeek V4-Flash's $0.14/1M/1M — roughly 5% less per token all in. The best dollar-for-dollar coding model for agentic pipelines that doesn't need to do anything else. Compare it first if DeepSeek V4-Flash's pricing is the thing stopping you.
What is a faster alternative to DeepSeek V4-Flash?
DeepSeek V4-Pro — balanced against DeepSeek V4-Flash's fast, with 1M tokens of context. Worth the swap when response time is what your users notice rather than the last few points of reasoning depth.
Newsletter
Get notified when DeepSeek V4-Flash pricing changes
We track pricing daily. When this model drops or spikes, you'll know first.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.