by DeepSeek

DeepSeek-V4 Flash review, pricing and verdict

DeepSeek's April 2026 open-weight Flash model, retired on 10 September 2026 and replaced by DeepSeek V4.1 Flash.

Best for
Self-hosting the original April weights
Not for
New API projects
  • Retired
  • Open source

Should you use DeepSeek-V4 Flash?

DeepSeek retired V4 Flash on 10 September 2026. Its API name now routes to V4.1 Flash, billed at $0.15 input and $0.60 output per million tokens off-peak. Use deepseek-flash for new work. The open MIT weights stay on Hugging Face for teams that need the original April model, which scored 79% on SWE-bench Verified.

Best for

  • Self-hosting the original April weights The MIT-licensed weights stay on Hugging Face, so a pinned April model can still be run without DeepSeek's API.
  • Understanding legacy deepseek-v4-flash calls The old name now serves V4.1 Flash, so this page explains what changed in price and behaviour.

Not for

  • New API projects DeepSeek retired the model and calls deepseek-flash the supported name; the legacy routing is described as temporary.
  • Reproducing April results through the API The alias no longer serves the April model, so outputs can differ.

DeepSeek-V4 Flash pricing

Input, per 1M tokens
$0.15/1M
Output, per 1M tokens
$0.60/1M
Cached input, per 1M tokens
$0.00/1M
Pricing model
per-token
Free tier
No

What a real job costs

JobInputOutputTotal
Summarise a 20-page PDF$0.0045$0.0006$0.0051
Support reply$0.0003$0.0002$0.0005
One coding agent run$0.030$0.012$0.042

Budgets: 20-page PDF = 30k in / 1k out · Support reply = 2k in / 300 out · Coding agent run = 200k in / 20k out. Computed from the vendor's per-token prices at render time; cached-input discounts are not applied.

Retired by DeepSeek. Calls to the legacy name are billed at the current Flash rates listed in the plan tiers, with peak rates twice the off-peak ones. The vendor's pricing page (read 2026-11-09) lists no free tier. The replacement model is priced the same.

Self-hosting requirements

Min vram gb
33
Self-hosted
Yes
Recommended GPU
1x H100 80GB at FP8, or 2x H100 for high throughput

What DeepSeek-V4 Flash can do

DeepSeek-V4 Flash, the 284B-parameter open-weight model released in April 2026, was retired by DeepSeek on 10 September 2026. Its legacy API name now serves DeepSeek V4.1 Flash with a 1M-token context window, and the MIT-licensed April weights remain available on Hugging Face for self-hosting.

What goes in and what comes out

Reads

  • text
  • tool-calls

Writes

  • text
  • tool-calls

Capabilities

  • Function callingSupported
  • Structured outputSupported
  • Tool useSupported
  • VisionSupported
  • Audio inputNot supported
  • Audio outputNot supported
  • Code executionNot supported
  • Video inputNot supported
  • Web browsingNot supported
Reasoning modes
standard, high, max
Long context recall
high

Key features

  • Retired and routed to its successor

    DeepSeek retired the model; its legacy API name now routes to the replacement and bills at Flash rates.

  • 1M-token context

    The 1,048,576-token window with a 384,000-token output cap carries over to the replacement model.

  • Open MIT weights

    The April 2026 model, 284B parameters with 13B active per token, is published on Hugging Face under an MIT licence for self-hosting and fine-tuning.

  • Peak and off-peak pricing

    Off-peak and peak rates apply; weekday-morning UTC peak hours cost twice as much as every other hour.

Known limitations

  • Retired by DeepSeek on 10 September 2026; the API alias now serves a different model.
  • Text-only at launch; image input arrived with V4.1 Flash.
  • DeepSeek's pricing page lists no free tier for the Flash model.

Who it is for

  • Startups and cost-sensitive teams building agentic systems or autonomous research platforms
  • Open-source-first developers and researchers requiring reproducible, deployable models
  • Enterprises with long-context use cases (contract review, document analysis, research synthesis) and tight budgets
  • Organizations avoiding US-based cloud providers or requiring self-hosted ML infrastructure

Supported languages

  • English
  • Spanish
  • French
  • German
  • Italian
  • Portuguese
  • Dutch
  • Russian
  • Japanese
  • Korean
  • Chinese
  • Arabic

DeepSeek-V4 Flash benchmarks

DeepSeek-V4 Flash against its nearest rivals

SWE-bench Verified % solved, higher is better

GPQA Diamond % correct, higher is better

Output speed tokens per second, higher is better

Where it sits

Cheaper than 75% of the 107 GA models with a published price, mid-pack on SWE-bench Verified (rank 17 of 40), and one of 42 that document a zero-data-retention option. Ranked against GA models; this record is not GA.

MeasureThis modelRankMedian
Blended price (3:1)$ per 1M tokens$0.262/M#24 / 107peer median $1.71/M
Output speedtokens/s94 tok/s#31 / 60peer median 94 tok/s
SWE-bench Verified% solved79%#17 / 40peer median 76.8%
GPQA Diamond% correct88.1%#32 / 64peer median 87.9%

Ranks are against GA models on Hokai that publish the same figure; ties share a rank.

Benchmark scores

  • HumanEval vendor-reported · 24 Apr 202696.4%Small programs that must pass hidden tests, % passing.
  • LiveBench vendor-reported · 24 Apr 202691.6%A rolling set of fresh questions that cannot have been in training data, % correct.
  • GPQA Diamond vendor-reported · 24 Apr 202688.1%PhD-level science questions that are hard to search for, % correct.
  • SWE-bench Verified vendor-reported · 24 Apr 202679%Real GitHub issues fixed end to end, % solved.
  • AA Intelligence Index cited: Artificial Analysis · 15 May 202647Composite of the evaluations Artificial Analysis runs for its index, 0 to 100.

Measured figures

LMArena Elo
1300
LMArena rank
#50
AA blended price
$0.21/M
Output speed
94 tok/s

A benchmark is an exam, not the job. Scores transfer unevenly between tasks, so weigh the one closest to your workload and read every figure with its source.

Safety benchmarks

Jailbreak resistance
moderate

Strengths and limits

Strengths

  • Open MIT weights let a team keep running the exact April model without DeepSeek's API.
  • The legacy API name keeps working for now, so existing integrations do not break overnight.
  • Migration is one configuration change, because deepseek-flash uses the same endpoint and price.
  • Open MIT weights remain available on Hugging Face, so self-hosting the April model is still possible.
Show 2 more
  • At release it scored 79% on SWE-bench Verified, 88.1% on GPQA Diamond and 96.4% on HumanEval per the model card.
  • Its 1M-token context carries over to the replacement model.

Limits

  • The model is retired, so no fixes or updates will arrive for the April release.
  • The legacy name now serves the replacement model, so behaviour can differ from what you tested.
  • DeepSeek calls the alias routing temporary, so it can be removed without a new release.

DeepSeek-V4 Flash alternatives

ModelLabContextIn / out per 1MReads and writes
DeepSeek-V4 Flash This pageDeepSeek1M$0.15 / $0.60text → text
DeepSeek-V4.1-FlashV4.1 Flash is the replacement the vendor routes the old name to, and it adds native image input. Pick it for any new work; V4 Flash only makes sense as self-hosted weights.DeepSeek1M$0.15 / $0.60text, image → text
DeepSeek-V4-ProV4 Pro stays on sale at $0.66 input and $1.98 output off-peak, roughly four times Flash's output price. Choose it when reasoning headroom matters more than cost.DeepSeek1M$1.32 / $3.96text → text
Qwen3.7-PlusQwen3.7-Plus is the non-DeepSeek option in the same long-context class, with built-in vision and video understanding.Alibaba Cloud1M$0.40 / $1.60text, image, video → text

DeepSeek-V4 Flash specs

Context window
1.0M tokens
Max output
384K tokens
Latency (P50)
1160 ms
Tokens per second
93.9 tok/s
Architecture
Mixture of Experts (MoE) with Hybrid Attention
Mixture of experts
Yes
Parameters
284B total / 13B active
VRAM requirements
33GB (quantized Q4_K_M), 80GB (FP8), 160GB+ (BF16 full weights)

Training data

Pre-trained on 32 trillion tokens of curated web text, licensed datasets, and synthetic reasoning traces. Post-trained with two-stage pipeline: domain-expert SFT + GRPO, unified via on-policy distillation.

Weight formats

  • safetensors
  • GGUF (community quantizations)

API endpoints

NameEndpointAuth
DeepSeek API (legacy name routes to V4.1 Flash)https://api.deepseek.comAPI Key

SDKs

  • Python
  • TypeScript
  • JavaScript
  • Java
  • Go
  • Rust

Prompting patterns

Show 2 prompting patterns
  • Explicit reasoning budget in non-thinking mode

    Set reasoning='high' or 'max' for complex tasks; use standard for routine queries.

    Thinking modes add visible chain-of-thought; high/max improve accuracy on STEM by 3-5 points but cost 2-3x more tokens.

  • XML tags for structured input

    <context>...</context><task>...</task>

    Like Claude, V4 Flash follows XML structure more reliably than markdown for complex prompts.

Known quirks

  • Reasoning modes have different output styles

    Workaround: Standard mode is concise; max mode is verbose. Test both for your use case.

  • Cache key resets if system prompt changes

    Workaround: Fix system prompts early; changing them mid-deployment resets cache savings.

DeepSeek-V4 Flash facts

Deployment
API · Self-hosted
Trains on your data
No
Licence
MIT License · Open source
Safety posture
Permissive
Released
24 Apr 2026
Current version
deepseek-v4-flash-20260424
Deprecated from
10 Sep 2026
Family
DeepSeek-V4
Variant
Chat
Data retention
API inputs retained 30 days for abuse monitoring and optional content moderation; deleted unless flagged. Zero-retention option available on enterprise plans.

Content restrictions

  • Refuses CSAM, bioweapons instructions, and direct malware code generation.
  • Generally more permissive on edge cases and requests than Claude or GPT; lighter constitutional alignment.

Safety controls

Rlhf
Yes
Red teaming
not publicly disclosed
System prompt
Yes
Constitutional AI
No
Content filter API
No
Moderation endpoint
No

Data governance

Iso 27001
No
Soc2 type2
No
GDPR compliant
No
HIPAA eligible
No
Trains on inputs
No
Eu AI act classification
General-purpose AI system
Data residency options
China (default), US (via AWS Bedrock), EU (via Azure)

Version history

  • deepseek-v4-flash-20260424 · 24 Apr 2026

    Initial V4 Flash release. 284B total and 13B active parameters, 1M context.

  • deepseek-v4.1-flash · 10 Sep 2026

    Successor. V4 Flash retired the same day; legacy names routed to V4.1 Flash.

  • deepseek-v4.1-flash-1017 · 17 Oct 2026

    V4.1 Flash upgraded to the 1017 build with API method and price unchanged.

What changed

  • Release DeepSeek-V4 Flash Public Preview (April 24, 2026) (deepseek-v4-flash-20260424) source

Questions about DeepSeek-V4 Flash

Is DeepSeek-V4 Flash still available?

Not as its own model. DeepSeek retired V4 Flash on 10 September 2026 when it released V4.1 Flash. The legacy name deepseek-v4-flash is still accepted, but it now runs the current Flash model at Flash rates, per DeepSeek's pricing page. The MIT-licensed weights remain on Hugging Face for self-hosting.

What does DeepSeek-V4 Flash cost on the API now?

Billing follows the current Flash rate card. Input costs $0.15 per million tokens on a cache miss and a fraction of a cent on a cache hit, and output costs $0.60 per million, all at off-peak times. Peak pricing is double on weekday mornings UTC. DeepSeek's pricing page lists no free tier.

What replaces DeepSeek-V4 Flash?

DeepSeek V4.1 Flash is the vendor's replacement, called as deepseek-flash. It keeps the 1M-token context and adds native image input. For more reasoning headroom at a higher price, DeepSeek V4 Pro is the larger sibling. Outside DeepSeek, Qwen3.7-Plus offers a similar long-context class with built-in vision.

Why did the model behind my deepseek-v4-flash calls change?

DeepSeek's changelog says the old names are temporarily routed to V4.1 Flash after the retirement of V4 Flash. Outputs, latency and token usage can therefore differ from what you saw in the spring. Switch your configuration to deepseek-flash and re-run your own evaluation, because the vendor calls the routing temporary and may remove it.

Can I still run DeepSeek-V4 Flash myself?

Yes. DeepSeek's April 2026 weights are on Hugging Face and MIT-licensed, so you can self-host or fine-tune them without the API. It is a Mixture-of-Experts model with 13 billion of its 284 billion parameters active per token. Self-hosting is the only way to get the original April behaviour now that the API alias points at a newer model.

About DeepSeek-V4 Flash

DeepSeek-V4 Flash was an open-weight Mixture-of-Experts language model that DeepSeek released on 24 April 2026 as the cheaper sibling of DeepSeek V4 Pro. DeepSeek retired it on 10 September 2026, when it shipped DeepSeek V4.1 Flash. The vendor's changelog says the old model names are temporarily routed to V4.1 Flash, and its pricing page (read on 9 November 2026) says requests to the legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are served by DeepSeek-V4.1-Flash-1017 and billed at the Flash price. New projects should call deepseek-flash directly. This page is kept as a record of what V4 Flash was and how to migrate off it.

Read the full background

At launch the model had 284 billion total parameters with 13 billion active per token, a 1,048,576-token context window and a 384,000-token output cap, under an MIT licence. The model card on Hugging Face listed 79% on SWE-bench Verified, 88.1% on GPQA Diamond and 96.4% on HumanEval, and Artificial Analysis gave it an Intelligence Index of 47. Those scores belong to the retired April model and say nothing about what the alias serves today. The open weights are still on Hugging Face, so a team that pinned V4 Flash for reproducibility can still self-host it.

What changed for API users is the model behind the name and the price. The current Flash model takes image input natively, which V4 Flash did not, and the 17 October update (DeepSeek-V4.1-Flash-1017) kept the API method and price unchanged. Prices per million tokens are now $0.15 input on a cache miss, $0.003 on a cache hit and $0.60 output at off-peak rates, with peak rates exactly double. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, excluding Chinese public holidays. At launch the model cost $0.14 input and $0.28 output per million tokens, and DeepSeek's pricing page lists no free tier for the Flash model today.

The case for staying on the retired alias is thin. It protects nobody from change, because the name already points at a different model, and DeepSeek called the routing temporary. Anyone who needs the exact April behaviour has to self-host the weights. Teams that wanted a larger step up in reasoning should compare DeepSeek V4 Pro, and teams that need a different vendor can look at Qwen3.7-Plus or GPT-4o mini. The vision preview that shared this family, DeepSeek V4 Flash Vision Exp, was retired in the same announcement. Vendor details sit on the DeepSeek company page, and every retired model is listed under retired models.

Sources and guides

More AI Models on Hokai

  • Hy4 flashTencent's Hy4 flash, released 16 November 2026, is a lower-cost multimodal model that reads text and images across a 1M-token window.
  • Nemotron-Labs-SigmaNVIDIA's Nemotron-Labs-Sigma, released in November 2026 in 8B and 3B sizes, writes text by denoising whole blocks of tokens in parallel, not one at a time.
  • Qwen4 MaxAlibaba's Qwen4 Max, released 16 November 2026, is a sparse mixture-of-experts model with a 1M-token context and text, image and video input.
  • Gemini 4 HydrogenGoogle DeepMind's November 2026 Gemini 4 model for coding agents, with four thinking levels from low to xhigh and a free tier in AI Studio.
  • Qwen4-Max-PreviewAlibaba's November 2026 Qwen 4 flagship preview, open only to Token Plan subscribers, with three reasoning levels from low to xhigh.
  • Claude Fable 5Superseded by Fable 5.1 and Fable 5.2 but still sold on the Claude API: Anthropic's first generally available Mythos-class model, released June 2026, with a 1M-token context.