← the radar · ✳ model tracker

DeepSeek V4.1 Flash

● live

DeepSeekopen weightsChinese labs stream

Released: September 10, 2026
Bench index: 68 · #11 of 45

The watchers keep tracking it: lineup changes, pricing moves and deprecations all alert.

Launch details

Live — Sep 10; 1M ctx; $0.15/$0.60 per 1M as our sweep read it, which is DeepSeek's off-peak rate (peak hours on weekdays cost double, $0.30/$1.20, per its pricing page); DeepSeek serves it under the unversioned id deepseek-flash, so the id itself never carries the version; Oct 17: upgraded in place to the 1017 checkpoint (DeepSeek-V4.1-Flash-1017, per DeepSeek's changelog), same deepseek-flash id and price, web and app updated too; open weights on Hugging Face as deepseek-ai/DeepSeek-V4.1-Flash-1017 (MIT, same config.json as the Sep 10 weights); OpenRouter lists it as deepseek/deepseek-v4.1-flash-1017 at $0.15/$0.60, 1M ctx, 384K max output; Oct 19: the Vals Index (updated Oct 18) scores the 1017 build 56.58%, #14 of 50 and the top open-weight model there, up from 51.32% for the Sep 10 build; $0.64 per test on Vals' run, against $0.33 for the Sep 10 build; Oct 19: DHS ordered DeepSeek products, and third-party products built on DeepSeek, excluded from and removed from US federal civilian agencies (Federal Register 2026-21447, publishing Oct 20); Oct 20: DeepSeek open-sourced DeepInfer (Apache-2.0), the inference engine it serves V4.1 Flash with, for NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs; Oct 20: Vals says the 1017 build completed 16.3 of 21 objectives on its coming offensive-cybersecurity benchmark, against 15.7 for GPT-6 Sol, on several times the tokens (a first finding; the full evaluation is due in the coming weeks).

Answering in 3.3s· answered 3h agoIs DeepSeek down? →

Watch for DeepSeek’s next model, free →Launch report (PDF)

Timeline

2026-10-20

DeepSeek open-sources DeepInfer, the engine behind V4.1 Flash, built for NVIDIA Blackwell and Huawei Ascend 950post

DeepSeek released DeepInfer, the inference framework it says serves DeepSeek-V4.1-Flash at scale, as open source on October 20 under the Apache 2.0 license.

2026-10-20

Vals: DeepSeek V4.1 Flash beat GPT-6 Sol on its offensive cybersecurity task, using about five times the tokenspost

Vals AI says DeepSeek V4.1 Flash 1017, an open-weight model, completed more objectives than OpenAI's GPT-6 Sol on a multi-step offensive cybersecurity task built on simulated corporate networks.

2026-10-19

Vals Index: DeepSeek's V4.1 Flash 1017 build is the top open-weight model, 56.58% at $0.64 a testpost

Vals AI now ranks DeepSeek-V4.1-Flash-1017 first among open-weight models on the Vals Index. It scores 56.58%, #14 of the 50 models on the index as updated October 18. Vals posted the result on X at 05:52 UTC on October 19.

2026-10-17

DeepSeek upgrades V4.1 Flash to the 1017 build: same price, higher scores on its own benchmarkspost

DeepSeek upgraded V4.1 Flash to a new checkpoint, DeepSeek-V4.1-Flash-1017, on October 17. DeepSeek's changelog says the API id and the price do not change: deepseek-flash now serves the new model, and DeepSeek's web and mobile apps have switched to it too.

2026-10-08

CrowdStrike: an open-source AI pentest agent running DeepSeek V4.1 Flash hit South Korean bankspost

CrowdStrike says an open-source AI penetration-testing agent called ARTEX, running on DeepSeek V4.1 Flash, was behind the data breaches at South Korean financial organizations from late September to early October.

2026-09-10

DeepSeek V4.1 Flash verified live — from the vendor's own announcement and catalog listingverified

2026-09-10

New on the radar: deepseek-flash appeared in models.dev (opencode-go)confirmed

On the bench index

LMArena: 1,665LiveBench: 72.9SimpleBench: 66.7Vals AI: 61.3Artificial Analysis: 32.5AW Index: 68

full comparison table →

Coding leaderboard

#9 of 45 on the coding index, at 76.1 · combines LMArena · Coding, LiveBench · Coding, Vals Vibe Code Bench

the AI coding leaderboard →

Open weights

#1 of 16 open-weights models on the Bench Index · deepseek-ai/DeepSeek-V4.1-Flash-1017 on Hugging Face · 763B parameters · license MIT

the open-weights leaderboard →

The official word

We alert on verified API behavior, never on announcements alone — but when DeepSeek speaks, it's here: deepseek news ↗

Where to use DeepSeek V4.1 Flash

80 providers list DeepSeek V4.1 Flash in the models.dev catalog (as of 2026-11-05), under the model ids shown below.

DeepSeekdeepseek-flash

Azuredeepseek-v4.1-flash

Azure Cognitive Servicesdeepseek-v4.1-flash

OpenRouterdeepseek/deepseek-v4.1-flash · deepseek/deepseek-v4.1-flash-1017 · ~deepseek/deepseek-flash-latest

Vercel AI Gatewaydeepseek/deepseek-v4.1-flash · deepseek/deepseek-v4.1-flash-1017 · deepseek/deepseek-v4.1-flash-1017-fast

302.AIdeepseek-flash

above.devdeepseek-v4.1-flash

AI/ML APIdeepseek/deepseek-v4.1-flash

ai&deepseek-ai/deepseek-v4.1-flash-1017

AIHubMixdeepseek-v4.1-flash

AKI.IOdeepseek-v4.1-flash-552b

Alibaba (China)deepseek-v4.1-flash

Also listed by Alibaba Token Plan, Alibaba Token Plan (China), AMD, Arcee, Baseten, Boundless, Charm Hyper, Chutes, ClinePass, Cloudflare Workers AI, CommandCode, CoralBricks, CoreWeave, Cortecs, CrossModel, Deep Infra, DevPass (LLM Gateway), DigitalOcean, Eden AI, EmpirioLabs AI, engy, Fireworks AI, GreenPT, Hopper, Hugging Face, Inco, IteraCompute, Kenari, Kilo Gateway, LLM Gateway, Melious, Merge Gateway, Mistral, Modal, Modular, Munito AI, NanoGPT, NEAR AI Cloud, Nebius Token Factory, Neon, Neuralwatt, Novita AI, Nvidia, Ofox, Ollama Cloud, OpenCode, OpenCode Go, Prime Inference, Requesty, Scaleway, SCNet Token Plan, SenseNova (China), Standard Compute, Subconscious, Synthetic, Tempr Gateway, TensorX, Thinking Machines, Tinfoil, Together AI, Umans AI, Umans AI Coding Plan, Vancine, Venice AI, Vivgrid, Vultr, Wally, ZenMux.

Listings cover the model family, so a provider may list a sibling tier under the same name. Check the id before you build on it.

DeepSeek V4.1 Flash: quick answers

Is DeepSeek V4.1 Flash out?

Yes. DeepSeek V4.1 Flash is live and has been since September 10, 2026.

When was DeepSeek V4.1 Flash released?

September 10, 2026.

Is DeepSeek V4.1 Flash good for coding?

On the ArtificialWatch coding index it scores 76.1, #9 of 45 ranked models. The index combines LMArena · Coding, LiveBench · Coding and Vals Vibe Code Bench, taking the median of each model's percentile across them.

Is DeepSeek V4.1 Flash open source?

It has open weights: the trained model is public on Hugging Face as deepseek-ai/DeepSeek-V4.1-Flash-1017, 763B parameters as Hugging Face counts them, license MIT. Among the 16 open-weights models the ArtificialWatch Bench Index ranks, it is #1.

Who makes DeepSeek V4.1 Flash?

DeepSeek. ArtificialWatch also tracks DeepSeek V4, DeepSeek Code, DeepSeek R3, DeepSeek V4.1 Pro, DeepSeek V4.2, DeepSeek V5 and Next DeepSeek Flash from DeepSeek.

Where can I use DeepSeek V4.1 Flash?

80 providers list it in the models.dev catalog, including DeepSeek, Azure, Azure Cognitive Services, OpenRouter and Vercel AI Gateway. The exact model ids each one lists are on this page.