AI Model Guide

What Is DeepSeek V4.1 Flash in 2026? The New Architecture Explained for Business Teams

DeepSeek's September 2026 release is the smallest model in a new architecture family, and it is being positioned to replace both the previous Flash and the previous Pro. This is the plain-language read: what shipped, why the cache numbers matter more than the parameter count, and what got retired.

Distk Editorial Sep 2026 12 min read

DeepSeek V4.1 Flash is a 552B-parameter Mixture-of-Experts model released on 10 September 2026 with native image understanding, a 1M-token context window and up to 384K output tokens. Its new Causal Encoder-Decoder design activates only 8B parameters to read input and 16B to generate output, and its KV cache footprint is roughly a quarter of DeepSeek V4 Flash. That compression is why DeepSeek can bill cache-hit input at 0.003 US dollars per million tokens off-peak. The model is available as deepseek-flash through OpenAI-format and Anthropic-format endpoints, the weights are on Hugging Face under the MIT licence, and V4 Flash and V4 Flash Vision Exp are retired. DeepSeek announced V4 Pro would be routed to V4.1 Flash from 14 September, then said a day later that V4 Pro service continues with billing unchanged. For business teams in 2026, this is a model that is cheap because of engineering, not because of a promotion.

What Is DeepSeek V4.1 Flash in 2026?

DeepSeek V4.1 Flash is a multimodal Mixture-of-Experts language model released by DeepSeek on 10 September 2026, described by the company as the smallest model in its new architecture family and as smarter, faster and more efficient than what came before. It has 552B backbone parameters, processes images and text natively, supports contexts up to one million tokens, and generates up to 384K output tokens. It is served through the DeepSeek API under the model name deepseek-flash and released as open weights on Hugging Face under the MIT licence.

The word Flash is doing less work here than it does in Google's lineup. In Google's naming, Flash is the workhorse tier below Pro. In DeepSeek's September 2026 lineup, V4.1 Flash is the only model of its generation, and DeepSeek's own statement is that tests by multiple parties put it ahead of DeepSeek V4 Pro on performance, cost, speed and total runtime. A V4.1 Pro is referenced as forthcoming. Until it arrives, V4.1 Flash is both the small model and the best model DeepSeek sells.

AttributeWhat DeepSeek published in 2026Why a business team should care
Released10 September 2026; new pricing effective 04:00 UTC the same dayThe same month Claude Fable 5.1 and GPT-6 Astra landed at 10 and 50 US dollars per million tokens.
Size552B MoE backbone; 8B active parameters for input, 16B for outputCost tracks active parameters, not total. This is why it is cheap to run.
ModalityText and images in, text out; vision trained in from the start of pre-trainingReceipts, screenshots, creatives and PDFs without a separate vision model.
Context1M input tokens; 384K maximum outputWhole archives in one call; very long deliverables in one response.
ReasoningThinking mode on by default; reasoning effort is an integer from 1 to 100A finer cost and quality dial than most vendors' three or four levels.
APIOpenAI-format and Anthropic-format endpoints; Responses API; JSON output; tool calls; concurrency limit 2,500Drop-in for most existing integrations. Five times the concurrency of V4 Pro.
LicenceMIT, weights on Hugging FaceSelf-hosting is legally simple; practically it is a data-centre project.
RetirementsV4 Flash and V4 Flash Vision Exp retired; legacy names route to V4.1 FlashExisting callers keep working, at Flash prices.

How Does the DeepSeek V4.1 Flash Architecture Work in Plain Language?

DeepSeek V4.1 Flash uses what the company calls a Causal Encoder-Decoder architecture: a 40-layer Transformer split into a 20-layer encoder that reads the input and a 20-layer decoder that writes the output. The memory the decoder needs about the input, the KV cache, is projected once from the encoder's final state rather than rebuilt by every decoder layer. The practical result is that reading a token activates 8B parameters and writing a token activates 16B, out of 552B in total.

For a non-specialist, the useful mental model is a reader and a writer sharing one set of notes. Older designs made every layer of the writer keep its own copy of the notes on the whole document. This design has the reader produce one compact set, and the writer works from that. Fewer copies means less memory per token, which means more requests fit on the same hardware, which means a lower price. DeepSeek's headline framing in 2026 is that cache-hit charges often account for a large share of agent costs, and compressing the cache cuts those costs.

The three numbers that explain the price

Other components DeepSeek names in 2026

The model card lists Single-Pass mHC for residual mixing, a 196B-parameter Engram conditional memory accessed sparsely by token lookup, DSpark speculative decoding for faster generation, one shared expert plus 384 routed experts per MoE layer with 6 routed experts active per token, and a from-scratch vision encoder called DeepSeek-ViT. None of these need to be understood to use the model. They are listed here so that a reader knows the terms exist when they appear in a vendor pitch.

What Changed From DeepSeek V4 Flash and V4 Pro in 2026?

Three things changed: the architecture, the training scale, and the lineup. V4.1 Flash was pre-trained from scratch on 45T tokens of multimodal data, with context extended to 1M tokens partway through training, and post-trained with what DeepSeek describes as large-scale automated synthesis of agent tasks and environments. The lineup change is the one with operational consequences, and the migration guide covers it in detail.

Model (per DeepSeek's model card)BackboneActive parametersVisionStatus in September 2026
DeepSeek V4 Flash284B13BSeparate Vision Exp modelRetired; name routes to V4.1 Flash
DeepSeek V4 Pro 08131.6T49BNot supportedPhase-out announced 10 Sep; continuation announced 11 Sep
DeepSeek V4.1 Flash552B8B input, 16B outputNativeGenerally available as deepseek-flash
DeepSeek V4.1 ProNot publishedNot publishedNot publishedReferenced as forthcoming; no date

On DeepSeek's own instruct benchmarks at maximum reasoning effort, V4.1 Flash scores above V4 Pro on most agentic rows: Terminal-Bench 2.1 at 90.6 versus 87.9, DeepSWE v1.1 at 74.2 versus 62.7, AutomationBench at 54.8 versus 43.2, and Agent's Last Exam at 31.8 versus 25.7. V4 Pro retains an edge on some knowledge rows in the base-model comparison, such as SimpleQA-Verified and LongBench-V2. The full picture, including where V4.1 Flash trails frontier models, is in the benchmark guide.

What Happened With the V4 Pro Retirement in 2026?

Two official DeepSeek pages say two different things, and both are current at the time of writing. The announcement of 10 September 2026 states that DeepSeek is phasing out V4 Pro and that from 04:00 UTC on 14 September all deepseek-v4-pro requests will route to V4.1 Flash at V4.1 Flash rates until V4.1 Pro launches. The API pricing page, updated 11 September, carries a footnote stating that in response to user demand DeepSeek has decided to continue providing API services for V4 Pro after 14 September with the billing method unchanged, and will give further notice of any change.

The sensible reading is that the pricing page is the later and more operational document, so V4 Pro stays callable for now at its own rates. The sensible planning posture is to treat V4 Pro as a model with a published end-of-life intent and no published date, and to have the V4.1 Flash migration tested and ready. This is not unusual for DeepSeek in 2026; the August rate card change was also announced with days of notice.

How Much Does DeepSeek V4.1 Flash Cost in 2026?

DeepSeek V4.1 Flash costs 0.30 US dollars per million cache-miss input tokens and 1.20 per million output tokens at peak, and half that off-peak, with cache-hit input at 0.006 peak and 0.003 off-peak. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays. Those rates put it at roughly one thirtieth to one fortieth of the list price of Claude Fable 5.1 or GPT-6 Astra on cache-miss input and output. The pricing guide works through what that means for an Indian working day and for a real workload; the September 2026 model comparison puts it alongside every other frontier rate card.

Where Is DeepSeek V4.1 Flash Available in 2026?

Why Does DeepSeek V4.1 Flash Matter for Business Teams in 2026?

Because it resets the floor. In the same month that the two Western flagships converged on 10 and 50 US dollars per million tokens, DeepSeek released a model that its own tables place close to or above the previous generation of those flagships on agentic benchmarks, at a price two orders of magnitude lower. Whether the benchmarks hold on your workload is a question only your evaluation set can answer. Whether the price is real is not in question; it is on the rate card.

We run DeepSeek in our own content engine, so this is a model we pay for rather than one we read about. The marketing team guide covers the use cases, the data-handling questions a China-served API raises under the DPDP Act and GDPR, and the self-hosting alternative the MIT licence makes possible.

What Are the Common Mistakes to Avoid in 2026?

Key Takeaways for 2026

Distk helps growth teams across India and internationally decide where a model like DeepSeek V4.1 Flash belongs in a routed stack, what to keep on a frontier tier, and how to handle the data questions before procurement asks them. If this release is on your 2026 shortlist, that assessment is where we start.

DeepSeek V4.1 Flash in 2026: FAQs

What is DeepSeek V4.1 Flash?

A 552B-parameter Mixture-of-Experts model released 10 September 2026 with native vision, 1M context and 384K output, using a Causal Encoder-Decoder design that activates 8B parameters for input and 16B for output. Served as deepseek-flash and released as MIT-licensed weights.

Is DeepSeek V4.1 Flash better than V4 Pro?

On DeepSeek's own agentic benchmarks, yes: Terminal-Bench 2.1 90.6 versus 87.9, DeepSWE 74.2 versus 62.7, AutomationBench 54.8 versus 43.2. V4 Pro's base model keeps an edge on some knowledge rows. DeepSeek states third parties found V4.1 Flash ahead on performance, cost, speed and runtime.

Is DeepSeek V4 Pro being retired in 2026?

The 10 September announcement said V4 Pro requests would route to V4.1 Flash from 14 September. The pricing page updated 11 September says V4 Pro service continues after 14 September with billing unchanged, in response to user demand. Treat it as end-of-life intent without a date.

What model name do I use for DeepSeek V4.1 Flash?

deepseek-flash. The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp still work and are served by V4.1 Flash at Flash prices.

Does DeepSeek V4.1 Flash support images?

Yes. Vision is native, trained in from the start of pre-training via a from-scratch DeepSeek-ViT encoder. The model card reports 95.6 on DocVQA. V4 Pro does not support vision.

Why is DeepSeek V4.1 Flash so cheap in 2026?

Because its KV cache is about 890 bytes per token, roughly a quarter of V4 Flash, needing a quarter of the GPU memory and an eighth of the SSD. More users fit on the same hardware, and DeepSeek says it is passing the savings on.

Know where the cheap model belongs before you route to it

Distk maps your workflows against DeepSeek V4.1 Flash and the frontier tiers, builds the evaluation set, and answers the data-handling questions before your client's procurement team asks them. In 2026, that is the work between a rate card and a decision.

Start the conversation →