Pull to refresh
Logo
What Changed
Following Why Sign Up
DeepSeek releases V4.1-Flash, a cheaper model for long-context AI agents

DeepSeek releases V4.1-Flash, a cheaper model for long-context AI agents

New Capabilities

New architecture cuts per-token memory to a quarter of the prior generation; a planned V4-Pro retirement was cancelled

September 17th, 2026: DeepSeek publishes V4.1-Flash technical report

Overview

Updated Oct 8

DeepSeek released V4.1-Flash on September 10, a model that reads a million tokens of context while using a fraction of the memory its predecessors needed. Its key-value cache, the memory a model keeps to recall what it has already read, drops to 890 bytes per token, one quarter of the prior Flash generation.

The price held: off-peak cached input runs $0.003 per million tokens. But the planned retirement of V4-Pro did not happen. DeepSeek said on September 11 it would keep serving V4-Pro with billing unchanged, after developers objected to a silent switch of the weights behind a production model name.

Why it matters

Off-peak cached input costs $0.003 per million tokens, making always-on AI agents cheap enough to run at scale.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

890 bytes/token
Key-value cache per token
Global cache footprint, one quarter of V4-Flash's and one eighth of its persistent SSD cache. Confirmed in the September 17 technical report.
1 million tokens
Context window
Native input context, with maximum output of 384,000 tokens.
$0.003
Off-peak cache-hit input price per million tokens
Half the peak rate. Off-peak cached input is one sixth the equivalent V4-Pro rate, which remained unchanged after the retirement reversal.
45 trillion
Training tokens
Multimodal corpus the model was trained on from scratch.

Voices

Curated perspectives — historical figures and your fellow readers.

Sign up to generate historical perspectives on this story.

Log in Sign Up

People Involved

Liang Wenfeng
Liang Wenfeng
Founder of DeepSeek
Runs DeepSeek from Hangzhou; rarely speaks publicly

Organizations Involved

DeepSeek
DeepSeek
AI lab
Released V4.1-Flash; kept V4-Pro after reversing its retirement plan

Chinese AI lab that releases open-weight models and runs a low-cost API.

Timeline

4 events Latest: September 17th, 2026 · 1 month ago
Tap a bar to jump to that date

  1. DeepSeek publishes V4.1-Flash technical report

    Latest Release

    The arXiv paper details the Causal Encoder-Decoder architecture, FP4 key-value caching, and the SWA Bounded Replay optimization behind the 890-byte-per-token cache.

  2. DeepSeek announces V4.1-Flash architecture details

    Announcement

    Details the causal encoder-decoder design, 890-byte-per-token FP4 cache, and September 14 V4-Pro retirement.

  3. DeepSeek cancels the V4-Pro retirement plan

    Migration

    DeepSeek drops its plan to route deepseek-v4-pro requests to V4.1-Flash from September 14. The notice on its platform says V4-Pro service continues after that date with billing unchanged, after developers objected to the planned swap.

  4. V4.1-Flash goes live on DeepSeek API

    Release

    Model launches as deepseek-flash with new peak and off-peak pricing. V4-Flash requests now route to it.

Scenarios

1

DeepSeek ships V4.1-Pro, extending its efficiency lead

Likely Resolves by Q1 2027

Discussed by: DeepSeek's own API announcement, which says V4-Pro will route to V4.1-Flash until V4.1-Pro launches

DeepSeek has said a V4.1-Pro is coming. If it pairs the causal encoder-decoder architecture with a larger backbone, the memory savings from this release would carry into a flagship model. Competing labs would then face a cost benchmark set by a Chinese model rather than by OpenAI or Anthropic.

2

Long-context failures surface, slowing enterprise adoption

Possible Resolves by Q2 2027

Discussed by: DeepSeek's model card, which warns sparse-selection errors and state reconstruction could degrade edge cases on long contexts and cache resumption

The model card admits the architecture's robustness limits are untested. Sparse attention can mis-select tokens over very long contexts, and the sliding-window state is reconstructed approximately rather than stored. Enterprises running long-lived agents could hit these edge cases, turning the cost savings into reliability problems and forcing DeepSeek to ship fixes.

3

Rivals match DeepSeek's cached-input pricing

Uncertain Resolves by Q2 2027

Discussed by: Dataconomy and The Register coverage noting the ultralow pricing targets agent workloads

At $0.003 per million cached input tokens off-peak, DeepSeek prices below the lowest tiers of OpenAI, Anthropic, and Google's flagship lines. If the price holds and quality checks out, competitors face pressure to cut cached-input rates for long-context agents, compressing margins across the industry.

4

V4-Pro stays on the price card and keeps a quality edge

Likely Resolves by End of 2026

Discussed by: DeepSeek's own multi-party testing, reported by OrcaRouter, which found V4-Pro retaining a quality edge on some production workloads where V4.1-Flash wins public evals

DeepSeek reversed the September 14 routing plan after developers objected to a silent swap behind the deepseek-v4-pro model name. The pricing page still lists V4-Pro-0813 at its higher rates. If V4-Pro keeps an edge on enterprise workloads, the two models coexist at different prices, and the V4.1-Flash savings apply only to users who opt in.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

January 2025

DeepSeek R1 market shock (January 2025)

DeepSeek released R1, a reasoning model trained for a reported fraction of what rivals spent, matching frontier performance on several benchmarks. On January 27, 2025, Nvidia lost about $589 billion in market value as investors questioned whether frontier AI required as many chips as projected.

Then

R1 topped app download charts worldwide and forced an industry-wide reassessment of training costs.

Now

Reset the assumption that frontier AI required billion-dollar training runs, and made efficiency a first-class goal for every lab.

Why this matters now

V4.1-Flash runs the same playbook, but targets inference memory cost rather than training cost.

May 2024

DeepSeek V2's Multi-head Latent Attention (May 2024)

DeepSeek-V2 introduced Multi-head Latent Attention, which compressed the key-value cache into a low-rank latent space. The technique cut cache memory sharply and let the model serve longer contexts on the same hardware.

Then

V2 cut inference costs and boosted throughput relative to comparable-size models of its day.

Now

MLA was widely copied across the industry and became the template for cache-efficient attention.

Why this matters now

V4.1-Flash's 890-byte-per-token cache is the direct descendant of that design, pushing the same obsession from training efficiency into serving efficiency.

February 2024

Google Gemini 1.5's million-token context (February 2024)

Google announced Gemini 1.5 Pro with a 1-million-token context window, the first mainstream model to handle that much input at once.

Then

The announcement reset expectations for what long-context AI could process, from entire books to full codebases.

Now

Million-token context became a benchmark that major labs now race to match; the differentiator shifted to how cheaply it can be served.

Why this matters now

DeepSeek's contribution is not the 1M window itself, but making that window affordable enough for continuous agent workloads.

Sources

(12)
+6
The Cosmic Meta Dataconomy DeepSeek API Docs Hugging Face The Register Tech Times Tao Media OrcaRouter Codersera SandBase Blog arXiv TokenSave