Skip to main content
AI Choice Engine

DeepSeek

DeepSeek V4.1 Flash

DeepSeek V4.1 Flash is a DeepSeek model with 1M context, Not verified in / Not verified out per million tokens, last verified October 25, 2026. Open weights: yes. Status: Retired since October 17, 2026; the current pick is DeepSeek V4.1 Flash 1017. Public-board scores are linked to the publisher; vendor-only figures stay in the watch-out, not on the board.

The September 2026 MIT-licensed DeepSeek V4.1 Flash checkpoint, retained separately from the October 17 API and weight update.

Recorded profile verification October 25, 2026Individual provider fields may changeAccess: retired — The September checkpoint is retained for native-weight and historical evaluation references. On October 17 the rolling deepseek-flash API moved to the separately catalogued Flash-1017 checkpoint; that live alias is not a pinned route to these September weights.

AI model specification details
SpecificationDeepSeek V4.1 Flash
ProviderDeepSeek
TierBalanced
Context window1M
Max output384K
Input / 1M tokensNot verifiedUnverified
Output / 1M tokensNot verifiedUnverified
WeightsOpen
Parameters552B MoE backbone + 196B Engram memory (causal encoder-decoder; 8B active prefill, 16B decode)
Reasoning levelslow, high, max
Modalitiestext, image
LicenseMIT
ReleasedSeptember 10, 2026

Pricing tiers: The original September checkpoint is no longer separately sold through the rolling first-party alias. Its MIT weights remain available for self-hosting; no current independent hosting tariff is verified here. Current rolling deepseek-flash pricing belongs to DeepSeek-V4.1-Flash-1017, not a new evaluation of these original weights.

Publisher evidence

Published benchmark results

Each result keeps its source date visible. Dates refer to source publication or capture unless the source explicitly states when it measured the result. A missing benchmark is not treated as a zero.

Published benchmark results for DeepSeek V4.1 Flash
BenchmarkScoreSource dateSource
Artificial Analysis Intelligence Index [max]DeepSeek V4.1 Flash (max)392026-10-01Artificial Analysis · View source
GPQA Diamond [max]Max effort; reasoning evaluation; complete run settings and experiment date unverifiedCross-model comparability not verified90.9%2026-10-02DeepSeek V4.1-Flash native model card (vendor-reported; October 2 source capture, not experiment date) · View source
DeepSWE 1.1 [max]mini-SWE; Max effort; N=8 samples/task; 1M context; temperature=1; top_p=0.95; max_steps=500; experiment date unverifiedCross-model comparability not verified74.2%2026-10-02DeepSeek V4.1-Flash native model card (vendor-reported; October 2 source capture, not experiment date) · View source
Terminal-Bench 2.1 [max]DeepSeek Harness Minimal; Max effort; N=3 samples/task; Linux containers; no network; 1M context; temperature=1; top_p=0.95; max_steps=500; experiment date unverifiedCross-model comparability not verified90.6%2026-10-02DeepSeek V4.1-Flash native model card (vendor-reported; October 2 source capture, not experiment date) · View source
Terminal-Bench 3.0 [max]DeepSeek Harness Minimal; Max effort; 1M context; temperature=1; top_p=0.95; sample count/task manifest and experiment date unverifiedCross-model comparability not verified30%2026-10-02DeepSeek V4.1-Flash native model card (vendor-reported; October 2 source capture, not experiment date) · View source
Terminal-Bench 4.0 [max]DeepSeek Harness Minimal; Max effort; 1M context; temperature=1; top_p=0.95; sample count/task manifest and experiment date unverifiedCross-model comparability not verified31.2%2026-10-02DeepSeek V4.1-Flash native model card (vendor-reported; October 2 source capture, not experiment date) · View source

Best for

  • Open-weight price-performance
  • High-volume hosted agents
  • Self-hosting with MIT weights

Watch out

Vendor task scores and the Artificial Analysis composite index measure different things. API low/high/max effort is separate from the checkpoint's 1–100 control; non-thinking API mode is also supported. Peak/off-peak billing follows time and Chinese public holidays. Responses and Anthropic-compatible APIs are supported; FIM is non-thinking only. Native prompt encoding maps low/high/max to 50/75/100 and defaults to high=75 in thinking mode; this does not establish the hosted API default. The native 552B backbone excludes 196B Engram conditional memory. Published minimal inference is a readable reference, not verified production serving; its uninitialized self-test is not real-model validation. The 384K output cap is a hosted API limit, not a confirmed native generation ceiling. This is the September checkpoint profile: hosted limits are historical reference values, while current rolling-API calls and prices belong to Flash-1017. No successor benchmark result is inherited.

History

DeepSeek V4.1 Flash timeline

Release and status events for this model that passed two-source verification — dated, sourced, and linked.

  • DeepSeek V4.1-FlashReleaseSeptember 10, 2026

    DeepSeek releases V4.1-Flash (MIT weights, 1M context, image input) as its new default at $0.30/$1.20 per MTok peak ($0.15/$0.60 off-peak) under the API id `deepseek-flash`, and says it beats V4-Pro on performance, cost and speed.

    DeepSeek — V4.1-Flash release notes

Common questions

DeepSeek V4.1 Flash

Answered from the verified figures on this page rather than general guidance.

What is DeepSeek V4.1 Flash's context window?
DeepSeek V4.1 Flash accepts about 1M tokens of context. That only matters if you routinely send very long documents, large codebases, or multi-turn histories that approach that limit.
What is DeepSeek V4.1 Flash best for?
DeepSeek V4.1 Flash is a balanced tier from DeepSeek. It suits open-weight price-performance, high-volume hosted agents, self-hosting with mit weights. Vendor task scores and the Artificial Analysis composite index measure different things. API low/high/max effort is separate from the checkpoint's 1–100 control; non-thinking API mode is also supported. Peak/off-peak billing follows time and Chinese public holidays. Responses and Anthropic-compatible APIs are supported; FIM is non-thinking only. Native prompt encoding maps low/high/max to 50/75/100 and defaults to high=75 in thinking mode; this does not establish the hosted API default. The native 552B backbone excludes 196B Engram conditional memory. Published minimal inference is a readable reference, not verified production serving; its uninitialized self-test is not real-model validation. The 384K output cap is a hosted API limit, not a confirmed native generation ceiling. This is the September checkpoint profile: hosted limits are historical reference values, while current rolling-API calls and prices belong to Flash-1017. No successor benchmark result is inherited.
Can I self-host DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash publishes open weights, but self-hosting depends on the licence, hardware footprint, quantisation quality, and serving stack. A hosted API is often cheaper until you have measured throughput and concurrency on your own hardware.