Project AURIONAI Use-Cases Research & Innovation
DeepSeek
multimodal LLM

DeepSeek-V4.1-Flash

DeepSeek open MoE model trained from scratch on 45T multimodal tokens.

DeepSeekDeepSeekChinaDeepSeek V4Family
Figure from the DeepSeek-V4.1-Flash model card.
Levelmodel
Classmultimodal LLM
Parameters552B
Context1,000,000
Released10 Sep 2026
Opennessopen weights permissive
Runs onmulti-GPU node
Statusactive

From the model's own page; who made each picture and where it comes from are listed on the Image Credits page.

Benchmark Results

vendor-reportedindependent
Arena Text leaderboardhuman preference arena
deepseek-v4.1-flash-maxindependent
1,475 ±7
1,2001,4001,600

Score in Elo; the axis starts at 1,200, as Elo has no zero. Higher is better.

rank 39, Arena Text overall8 Oct 2026 arena.ai
Artificial Analysis Intelligence Index v4.3.2composite index
maxindependent
39
03060

Score in index. Higher is better.

AA leaderboard, 205 current models10 Oct 2026 artificialanalysis.ai
Codeforces ratingcoding
vendor-reported
3,471
02,0004,000

Score in rating. Higher is better.

huggingface.co
DeepSWE v1.1coding
mini-SWEvendor-reported
74.2
050100

Score in %. Higher is better.

huggingface.co
GPQA Diamondknowledge and reasoning
vendor-reported
90.9
050100

Score in %. Higher is better.

huggingface.co
Terminal-Bench 2.1agents and tool use
vendor-reported
90.6
050100

Score in %. Higher is better.

huggingface.co
DeepSeek Harness Minimalvendor-reported
90.6
050100

Score in %. Higher is better.

huggingface.co
Terminal-Bench 4.0agents and tool use
vendor-reported
31.2
050100

Score in %. Higher is better.

huggingface.co
Terminal-Bench 3.0agents and tool use
vendor-reported
30
01530

Score. Higher is better.

huggingface.co
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

R&D Topics

Efficiency architectures (MoE, sparse attention, long context)produced byOpen-weight frontier releases as research substrateused by

Identity

Developer
DeepSeek
Family
DeepSeek V4
Released
10 Sep 2026 sourcessecondary report
Weights published
2 Oct 2026
Record checked
10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
MoE MoE "causal encoder-decoder" (20 + 20 layers), 384 routed experts (6 active + 1 shared), Engram memory module (196B), CSA2 attention
Total parameters
552B sources
Active parameters
16B
Checkpoint parameters
763B
Note
763B listed (552B backbone); 8B active prefill, 16B decode
Context length
1,000,000 tokens
Input
text, image
Output
text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open weights permissive
Licence
MIT License sources
Access routes
API, open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
multi-GPU node
Footprint
T3 estimate (unverified). HF updated about 2026-10-02.
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Lineage

The notes record no lineage for this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. deepseek-ai/DeepSeek-V4.1-Flash , Hugging Face huggingface.co https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash
  2. deepseek-ai (DeepSeek) huggingface.co https://huggingface.co/deepseek-ai
  3. LLM Leaderboard - Comparison of AI models from OpenAI, Anthropic, Google, SpaceXAI & others | Artificial Analysis artificialanalysis.ai https://artificialanalysis.ai/leaderboards/models
  4. Text Arena: LLM Rankings & AI Model Benchmarks arena.ai https://arena.ai/leaderboard/chat/text
  5. Best AI models, October 2026: Opus 5.5, GPT-6, Gemini slash-digital.io https://slash-digital.io/en/insights/ai-models-2026/
  6. Open-Source LLMs 2026: Which Are Actually Open-Licensed codersera.com https://codersera.com/blog/open-source-llms-landscape-2026/

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "DeepSeek-V4.1-Flash", https://aurion.semai.id/models/deepseek-deepseek-v4-1-flash/, records of October 2026.