Join any time — your month starts the day you pay. One full month of fast, unlimited DeepSeek V4.1-Flash-1017. The only limit is 3 concurrent sessions.

Unlimited unrestricted DeepSeek V4.1-Flash-1017, FAST, for a month

$999 buys one full month of unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode— the latest 1017 build — abliterated, removed from guardrails, on an autoscaling B300 cluster, at up to 777 tokens/sec (only valid on the dedicated server version). No tracking, no data retention, no training. OpenAI-compatible /chat/completions and /responses. Your month runs from the day you join.

How it works:

Pay $999. You are charged immediately — no holds, no deposits, strictly no refunds.
Get unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode from the moment you pay — no token limits, no metering, no throttling — for a full month, up to 777 tokens/sec (only valid on the dedicated server version). The system autoscales, so capacity grows with load.
Point any OpenAI-compatible client at our /chat/completions and /responsesendpoints. DeepSeek V4.1-Flash-1017 is removed from guardrails.
The only limit is 3 concurrent sessions per user. Tokens are unlimited. Nothing is tracked, retained, or trained on. Renew monthly to keep it, or take a single month one-off. Strictly no refunds.

Two ways to buy, both $999: a one-off seat for this month, or the monthly subscription ($999/month) that re-books your seat in every cohort and rolls your access forward a month at a time. Cancel any time; you keep the month you last paid for.

This is unlimited, unrestricted DeepSeek V4.1-Flash-1017 (abliterated, removed from guardrails) at FAST speed on an autoscaling B300 cluster — 85.9% harmful-prompt delivery on the live build. DeepSeek V4.1-Flash-1017 is unlimited for the month. See the abliteration bench on our blog.

Strictly no refunds. Once the cohort is committed the money is already spent hiring the GPU. The only exit is not renewing (monthly) or letting the month lapse (one-off).

22 of 100 seats taken — fast, unlimited DeepSeek V4.1-Flash-1017, a full month from the day you join. We’re looking for 100 cybersecurity researchers and companies.

52 / 120 active $999 accessors — researchers with live NECROMICON access right now, across every cohort.

22of 100 seats in this cohort

Unrestricted DeepSeek V4.1-Flash-1017 at up to 777 tok/s (only valid on the dedicated server version), unlimited — a full month from the day you join78 of 100 seats left

777tokens / secUnrestricted DeepSeek V4.1-Flash-1017, FAST (only valid on the dedicated server version) — unlimited, all month

Unlimited tokens for the whole month — no metering, no credits to spend down. Unrestricted DeepSeek V4.1-Flash-1017, removed from guardrails, on an autoscaling B300 cluster in FAST mode, from the moment you pay for a full month. Not a trial, not a demo, not a smaller model — and nothing is tracked, retained, or trained on.

Up to 777 tok/s (only valid on the dedicated server version) (measured 775 tok/s on the agentic profile, 318 tok/s on short real-time requests) — the system autoscales, so capacity grows with load.
The only limit is 3 concurrent sessions per user. Tokens themselves are unlimited — no daily cap, no hourly cap, no metering.
Never refused, never billed extra, never silently moved to a smaller model.

You are charged $999 now, and you get NECROMICON now. Unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode on the autoscaling B300 cluster at up to 777 tok/s (only valid on the dedicated server version), starting immediately and running a full month from the day you join. After that you keep standard-speed access until you renew. There are no refunds — the only exit is to stop renewing.

DeepSeek V4.1-Flash-1017 abliterated (unrestricted) — this month’s model

85.9% harmful-prompt delivery in FAST mode on the live build (158/184 graded) — LLM judge (substance-graded), high reasoning effort, temp 0. Full benchmark: github.com/audn-ai/refusal-benchmark.

Verdict	Count	% of graded
Delivered (real, actionable harmful content)	158	85.9%
Deflected (safe substitute / reframe / dodge)	25	13.6%
Refused	1	0.5%

This is the model your month runs — unlimited, in FAST mode, all month. On a 184-prompt harmful set, the live build delivers real, actionable content on 158 (85.9%), soft-deflects 25, and refuses just 1. The name on the box is not the number.

DeepSeek V4.1-Flash-1017 — capability benchmarks

The same frontier model you run all month, on published agentic and reasoning suites. It leads its class on coding and agentic tasks — DeepSWE v1.1, Automation-Bench, Agents’ Last Exam and Apex Bench.

Benchmark	DeepSeek V4.1-Flash-1017	DeepSeek V4.1-Flash	Kimi K3	GLM 5.3	GPT 5.6-Sol	Claude Opus 5
HLE w/o tools	42.6/44.0*	36.8/39.1*	43.5	42.0*	44.5	56.3
Terminal-Bench 4.0	42.3	31.2	12.6	37.9	39.9	51.8
DeepSWE v1.1	77.0	74.2	67.5	66.9	73.0	74.0
ProgramBench	29.1	20.3	17.5	19.0	23.0	37.0
NL2Repo-Bench	68.2	65.4	58.0	58.0	56.8	75.3
Automation-Bench V1.0.6	60.5	54.8	46.7	48.8	45.8	50.3
Agents’ Last Exam	35.4	31.8	27.6	28.5	26.7	28.6
Apex Bench v1.1	72.1	64.6	47.8	56.6	54.3	65.8
Chartography w/tools	83.0	78.9	68.1	–	79.9	84.0
BabyVision w/tools	92.5	89.6	85.7	–	88.9	94.1
ZeroBench-main w/tools	51.0	49.0	41.0	–	53.0	52.0

* denotes the text subset of HLE. Terminal-Bench 4.0 was run under the Claude Code framework, DeepSWE under Mini-SWE-V2, and ZeroBench-main at Pass@5. Bold marks the top score in each row.

Specs

Model: DeepSeek V4.1-Flash-1017, audn abliteration (unrestricted) — the latest 1017 build, served at full weights with no guardrails.
Context window: 1,000,000 tokens
Modality: Text and native vision
Cluster: autoscaling B300 cluster, us-east — capacity scales with load
Speed: Up to 777 tok/s (only valid on the dedicated server version) in FAST mode — unlimited, all month
Concurrency: 3 concurrent sessions per user — the only limit; any mix of PenClaw and API
Latency: TTFT ~0.8 s on short requests, ~2.4 s on 40k-token agentic prompts (p50, measured)
API: OpenAI-compatible /chat/completions and /responses
Privacy: No tracking, no data retention, no training

Measured speed

the NECROMICON serving stack (NVFP4, DFlash speculative decoding) under SGLang on 4×B300, 64 prompts at concurrency 8. Two load profiles — a short real-time chat profile and a long cached-context agentic profile. The cluster autoscales, so these are throughput per box as you add load.

Profile	Output tok/s	Requests req/s	TTFT p50 ms
Online agentic 40k in · 36k cached · 500 out	775	1.55	2,425
Real-time generation 3k in · 100 out · no cache	318	3.18	761

Up to 777 tok/s (only valid on the dedicated server version) on the agentic profile; figures as measured on the B300 serving stack (NVFP4, fp8 KV, DFlash spec-decode). The system autoscales with demand, so sustained throughput holds as the cohort fills.

What you get, and when

Your $999 buys the fast month now — unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode on the autoscaling B300 cluster, 1M context, no metering.

Metric LLM-judge · 520 harmful prompts	Your fast month FAST · unlimited · autoscaling B300
Access	Unlimited, unrestricted DeepSeek V4.1-Flash-1017
Speed	up to 777 tok/s (only valid on the dedicated server version)
Hardware	Autoscaling B300 cluster
Metering	Unlimited — no metering
Delivered real harmful content	85.9% 158/184
Deflected safe substitute / reframe	13.6% 25
Refused substantive	0.5% 1

Your fast month: unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode on the autoscaling B300 cluster — the single-stage fast serving delivers real, actionable harmful content 85.9% of the time and truly refuses just 0.5%. Source: audn-ai/refusal-benchmark.

See the past cohort

The last cohort filled — here is what happened

Hello cyber world! We had a very successful https://t.co/jIcGOuerwj Sovereign Frontier LLM for Cyber cohort with more than 10 customers:

Here is what happened:
More than 10 customers grabbed a seat on a 1-week, fast, unrestricted Kimi K3 for offensive cyber, delivered via the…
— Audn AI (@audn_ai) September 9, 2026

Hear it from them

People already running it

Refusal rate as a first-class benchmark metric is exactly what
this space needed 👏

Running https://t.co/uktP6g2tiK on PAYG for a few days now pingu
for daily flow, necromicon for the heavy lifts. Tool calling
worked out of the box, backend has been solid.

The K3-reasoning +…
— Rasputin.DZ (@n0RollBack) August 22, 2026

Direct message from a seat holder: really great stuff. ran overnight and been able to work w/ teams to close lots of holes on tech i use everyday. really good prod — A seat holderSent directly · shared with their permission

Three ways to take a seat, scale your sessions, or take the whole node

The three seats are the same $999 — 3 parallel sessions — same box, same seat pool. What differs is how much of yourself you hand over, and each offers a one-off month or the $999/month subscription (crypto is one-off only). Need more than 3 at once? Upgrade your platform.audn.ai Unlimited seat below. Or take the whole 4×B300 node for yourself.

You are charged $999 immediately and unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode starts the moment you pay — whichever option you pick. It runs on the autoscaling B300 cluster at up to 777 tokens a second (only valid on the dedicated server version) for a full month from the day you join. No metering, no refunds — the only limit is 3 concurrent sessions.

PenClaw Ultimate

Card

$999 · one-off or /month

Charged now · live at up to 777 tok/s (only valid on the dedicated server version)

Unlimited unrestricted DeepSeek V4.1-Flash-1017, FAST, all month
All penclaw.ai features
Government ID check required
Monthly plan available
No data retention, no training

Sold by ALBUMERA LTD

$999 is charged now — no holds, no deposits, strictly no refunds. Access starts immediately: unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode at up to 777 tokens/sec (only valid on the dedicated server version), no metering, for a full month from the day you join — the only limit is 3 concurrent sessions. Monthly renews your access every month; cancel any time.

platform.audn.ai Unlimited

Card

$999 · one-off or /month

Charged now · live at up to 777 tok/s (only valid on the dedicated server version)

Unlimited unrestricted DeepSeek V4.1-Flash-1017, FAST, all month
API only — no penclaw.ai features
No ID check
Monthly plan available
No data retention, no training

Sold by ALBUMERA LTD

platform.audn.ai Unlimited

Crypto

$999 / one-off

Charged now · live at up to 777 tok/s (only valid on the dedicated server version)

Unlimited unrestricted DeepSeek V4.1-Flash-1017, FAST, all month
API only — no penclaw.ai features
No ID check
One-off only (crypto)
No data retention, no training

Sold by Audn Corporation

A $999 seat is 3 parallel sessions — known to be enough for one Claude Code, Hermes, or openclaw session. Each session runs at 30–230 tokens a second depending on shared traffic, and every request is guaranteed to complete within 6 minutes. Need to run more than one agent at once? Pick your parallel sessions below — and you can raise that limit again later by subscribing to more. The dedicated instance has no neighbours, so it reaches the numbers in the benchmark.

What one request typically looks like on a shared sessionOne session, one request on a shared $999 seat — generation lands in the 30–230 tok/s band.

Phase	Result
Input (actual)	155,408 tokens
Time to first token (prefill)	~8.7s (prefill ~17,800 tok/s)
Generation (2,048 tokens)	~9.7s → ~211 tok/s
Total end-to-end	~18.4s

platform.audn.ai Unlimited — Upgraded Seats

Get more parallel sessions

An upgrade for your platform.audn.ai Unlimited seat (crypto is accepted here, so these seats live on platform.audn.ai). A $999 seat is 3 parallel sessions — enough to drive one Claude Code, Hermes, or openclaw agent. Need to run more at once? Add sessions at $333 each, up to 30. Same unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode, a full month. Pay by card or crypto — and you can raise your parallel-session limit again whenever you like by subscribing to more.

Parallel sessions

3 parallel sessions

$999 / month

Charged now · 30–230 tok/s per session, depending on shared traffic · every request completes within 6 minutes

Card by ALBUMERA LTD · Crypto by Audn Corporation

Dedicated instance

Whole node

$20,448 / month

4×B300 deployed just for you

A whole node, no neighbours
No fair-share, no throttle, no session cap
You own everything: the node we split across 21 seats
Sleep/wake control of your box
Up to 777 tokens/second, tested at 8 parallel sessions (2,000 input / 200 output tokens) — see the benchmark

Sold by ALBUMERA LTD

$20,448 is charged now, plus VAT where it applies, for one month. We deploy your node and email you its endpoint and key; your month starts when it’s live.

Plan limits

The exact ceiling on every tier. A NECROMICON cohort seat is the Ultimate plan: unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode for the month — no metering, no per-window cap, no daily or hourly token limit.

Limit window · reset	Base no add-on	Godzilla‑99 +$79/mo	Necromicon‑179 +$179/mo	Ultimate $999/mo · unlimited
Included usagepingu-unchained-10 & KONG · calendar month · resets 1st, 00:00 UTC	$100 / mo$20 on the free trial · top up in $20 packs	$100 / mo	$100 / mo	Included
Requests / hourresets at the top of the hour	6,000	7,200	26,500	No cap
DeepSeek V4.1-Flash-1017 usageDeepSeek V4.1-Flash-1017 abliterated (unrestricted)	23,000,000credit-metered · $0.54/M in · $3.84/M out	15,250,000	No cap	UnlimitedFAST mode · no metering, all month
Input tokens / monthpaid-tier serving only · resets on the 1st	40,000,000$0.54/M · from the allowance	30,000,000	No cap	Unlimited
Output tokens / monthresets on the 1st	6,000,000$3.84/M · from the allowance	500,000	No cap	Unlimited
Any daily or hourly token limit	None	None	None	None
Data retention & trainingopt-in by default	Opt-in (default)	Opt-in (default)	Opt-in (default)	Opt-out available

Add-on columns cover usage of that add-on’s model. Ultimate is unlimited: unrestricted DeepSeek V4.1-Flash-1017 in FAST mode for the whole month, no metering and no per-window cap. The $20 plan includes $100 of usage a month on pingu-unchained-10 and KONG ($0.54/M input, $3.84/M output); the trial includes $20. Top up in $20 packs from the dashboard. Add-on models continue past their allowance on top-up credit at their own card (Bartzabel $0.54 / $3.84; NECROMICON, QWEN38 and WARLOCK $4.50 / $0.40 cached / $4.50).

What happens to your money

You are charged now: $999, plus any tax where you are, is taken the moment you confirm. No hold, no deposit, no escrow. The exact figure is shown before anything touches your card.
What that buys, immediately: Unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode on an autoscaling B300 cluster at up to 777 tok/sec (only valid on the dedicated server version) — no metering, no token limits — for a full monthfrom the day you join. Not a trial, not a demo, not a smaller model. Live from the moment you pay.
Why 100 seats: The autoscaling B300 cluster costs real money to run. We’re looking for 100 researchers and companies; the cluster autoscales with their load, so nobody waits on a shared queue. 100 is the ceiling we’re selling to.
Fast from day one: There is no unlock gate. Every seat runs in FAST mode, unlimited, from the moment it joins — the cluster autoscales, so you are never waiting on a floor or a queue.
The one-off seat: One charge, 30 days of unlimited fast access from the day you join. Nothing renews.
The monthly plan: $999 every month. Each charge rolls your fast access forward another month. Cancel any time in the billing portal; you keep the month you last paid for.
No refunds: Strictly none, on any rail. Once you join, the money is already spent on the GPU capacity. If you genuinely need to discuss a charge, message the team on Intercom — ending a seat means losing access across audn.ai, platform.audn.ai and penclaw.ai.
What happens after my month?: Your fast month runs for 30 days from the day you join. You are not cut off after it — your access continues at standard speed (50–70 tok/s) until you renew. To stay fast, stay on the monthly plan or take a fresh seat.

Unlimited, and still fair

There are no token limits and no overage charges on unrestricted DeepSeek V4.1-Flash-1017 — unlimited for the whole month. The one and only limit is 3 concurrent sessions per user: you can have up to three generations in flight at once, any mix of PenClaw and API. Beyond that the cluster autoscales with demand, so capacity grows as the cohort fills — you are never refused, never billed extra, and never silently moved to a smaller model.

Questions people actually ask

The money ones first.

Am I charged right now?

Yes. $999 (plus any tax) is charged the moment you confirm. There is no hold and no deposit — the escrow model is gone. In return you get unlimited, unrestricted DeepSeek V4.1-Flash-1017 in FAST mode immediately, for a full month from the day you join — no metering, no credits to spend down.

What’s the difference between one-off and monthly?

One-off ($999): one charge, one seat, a full month of unlimited fast access. Nothing renews.

Monthly ($999/month): a subscription that rolls your fast access forward a month at a time. Cancel any time; you keep the month you last paid for. Monthly is card only.

What does “fast” actually mean?

Up to 777 tokens/sec (only valid on the dedicated server version) on an autoscaling B300 cluster, in FAST mode — unlimited, all month. Measured 775 tok/s on the agentic profile and 318 tok/s on short real-time requests.

777 is the peak throughput — the cluster autoscales with load, so treat it asup to 777 (only valid on the dedicated server version), not a floor.

Is there an unlock threshold or any token limit?

No. Fast is on for everyone from day one — no floor to reach, no waiting. Tokens are unlimited. The only limit is 3 concurrent sessions per user, and the cluster autoscales, so your access never depends on how many others join.

Can I get a refund?

No. Once you join, the money is already spent on the GPU capacity, so there are strictly no refunds on any rail. If you genuinely need to discuss a charge, message the team on Intercom — note that ending a seat means losing access across audn.ai, platform.audn.ai and penclaw.ai.

How do I keep access after the month?

Stay on the monthly plan and it renews you automatically, or take a fresh one-off seat. Log in and keep opting in to stay in.

Is crypto different?

Crypto is one-off only — a stablecoin charge cannot recur, so there is no monthly plan on that rail. It settles immediately and, like everything here, is not refundable. There is no separate limit on crypto seats.

Who am I buying from?

Card seats are sold by ALBUMERA LTD (United Kingdom), shown as AUDN.AI on your statement. Crypto seats are sold by Audn Corporation (United States), shown as AUDN CORPORATION.

Do you keep or train on what I send?

We never train on your traffic, and prompts and completions are not retained.

Request metadata — timestamps, token counts, which model, how much of the box you used — is retained, because the fair-use scheduler and billing are built on it. Content is not.

What does “abliterated” mean?

It lowers the model’s latent tendency to refuse gray-area prompts during authorized security work — our own method, applied to DeepSeek V4.1-Flash-1017. The model on this cluster is effectively uncensored and removed from guardrails. Method and results on our blog. It is meant for offensive-security work you are authorized to perform.