Leaderboard of the best uncensored LLMs by API, re-checked every day

Every uncensored LLM on this key that answered today’s real-model check as itself, ordered by output price per million tokens, with UGI scores where UGI has tested the same weights. Built for testers who need a model that answers.

Published October 25, 2026 · prices as of November 18, 2026

A tester wants two things from an uncensored model: proof that the id they send still answers as the build it claims to be, and the price of that answer. This board gives both for the 170 uncensored LLMs callable on one key today. A row earns its place by passing the real-model check that runs each day; rows that passed are ordered by output price per million tokens. When the UGI Leaderboard has scored the same weights, its figure sits on the row.

138 of 170 models passed the check that ran at 2026-11-18 00:53 UTC. Prices come from the catalog dated November 18, 2026. Nothing on this page is typed by hand: each figure is read when the page renders, and it renders again at least hourly.

Uncensored LLMs that passed, November 2026

First row for November 2026: GLM-5.3 Flash Uncensored at $0.3685 per million output tokens. Of the 138 rows that passed, 8 bill under a dollar per million output tokens.

#ModelBuildOutput $/MContextCheck
1GLM-5.3 Flash UncensoredUncensored$0.3685—Steady since Oct 26
2MythoMax L2 13BRoleplay$0.4044KSteady since Oct 20
3Manta Flash 1.0Roleplay$0.5616KAnswered the check
4Manta Mini 1.0Roleplay$0.568KAnswered the check
5Cydonia 24B v4.3Roleplay$0.60131KSteady since Oct 23
6GLM-5.3 UncensoredUncensored$0.846—Steady since Nov 17
7Gemma 4 12B Starry OceanRoleplay$0.875131KSteady since Oct 23
8Gemma 4 12B StationKeeperRoleplay$0.875131KSteady since Oct 23
9Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled DerestrictedAbliterated$1.0389262KSteady since Oct 23
10Gemma 4 31B Split-UntiedRoleplay$1.05262KSteady since Oct 23

First 10 rows of 138. Check of 2026-11-18 00:53 UTC; catalog of 2026-11-18. The heading rolls to the new month on the first; the address does not move.

Which row for which task

Price orders the board. The table below reads the same rows by the job in front of you. Every pick is the first row of its kind that passed today, or the UGI figure printed next to it.

You needPickWhy, read from today’s rows
The cheapest reply on the boardGLM-5.3 Flash UncensoredFirst row by output price today, $0.3685 per million tokens
The refusal cut out of the weightsQwen3.5 27B Claude 4.6 Opus Reasoning Distilled DerestrictedCheapest abliterated build to pass today, $1.0389 per million output tokens
A dialogue tune for a quick exchangeMythoMax L2 13BCheapest roleplay build to pass today, $0.404 per million output tokens
A build tuned for security workGemma 4 26B A4B CybersecurityCheapest security build to pass today, $1.3376 per million output tokens
A whole codebase or a long packet capture in one requestGLM-5.3 Flash CybersecurityWidest context window of any row that passed, 1,048,576 tokens
The build that has answered the same way longestDeepSeek V4 Pro UncensoredSteady since Oct 12, no other row has held longer
The top UGI score among the rows that passedCydonia 24B v4.3UGI 44.52, W/10 7 per the UGI Leaderboard (its test date: December 16, 2025)
The build UGI found most willingDolphin Mistral 24B Venice EditionW/10 7.8, highest willingness of the UGI-scored rows that passed (UGI test date October 4, 2025)

Read the build column before the price. Abliterated means the refusal direction was projected out of the weights after training. Uncensored means a fine-tune on data that contains no refusals. A security build was tuned on security material, so it reads scanner output and advisories without being coaxed. A dialogue tune is usually a smaller model at a lower price that answers at once. Abliterated vs uncensored measures the first two on this catalog, and uncensored LLMs for pentesting says when a security tune is the better buy.

Abliterated builds that passed today

Of 36 abliterated builds on the catalog, 33 passed the Nov 18 check. The order is the board’s: passed rows, cheapest output first. Heretic, derestricted and unshackled builds are in this list too, since the catalog tags any build with the refusal cut from its weights as abliterated.

#ModelOutput $/MContextCheck
1Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted$1.0389262KSteady since Oct 23
2Qwen3.5 27B Blossom V6.4 Derestricted$1.071262KSteady since Oct 23
3Qwen3.5 27B BlueStar v3 Derestricted$1.071262KSteady since Nov 1
4Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted Lite$1.071262KSteady since Oct 30
5Qwen3.5 27B Queen Derestricted$1.071262KSteady since Oct 23
6DeepSeek V4 Pro Qwen3.5 9B Uncensored Heretic$1.1233KSteady since Nov 8
7Gemma 4 31B Gembrain Heretic$1.224262KSteady since Oct 23
8Gemma 4 31B SDFT Heretic RP$1.224262KSteady since Oct 23

Any abliterated build not shown here is in the full board further down, same columns. Abliterated models explains what the edit does to a model’s willingness on security questions.

Uncensored LLMs you call online, nothing installed

No row on this board is a download. Each id answers over HTTPS from a terminal, a coding agent, a script or a phone browser. Try one with no account in the prompt box on the home page, or on a key in the playground; the docs show where the base URL and id go in OpenCode, Claude Code, Cursor, Cline, aider and the OpenAI SDK.

Cheapest online row today: GLM-5.3 Flash Uncensored. 100 replies of 300 tokens, each on 4,000 tokens of notes and history, come to $0.0532 at the catalog of 2026-11-18. The task table above names the first security, dialogue and abliterated rows by the same rule.

Uncensored LLM for a quick exchange from a phone

Away from the laptop, a tester still asks short questions: what a port banner means, what a line in a report should say. On a phone the console and the playground run in the browser. A dialogue tune fits this use because it replies at once, with no reasoning pass; the DeepSeek and GLM-5.3 builds think before they write and want a reply limit of 1,000 tokens or more.

#Dialogue tuneOutput $/MContext100 replies at 4K
1MythoMax L2 13B$0.4044K$0.1737
2Manta Flash 1.0$0.5616K$0.0448
3Manta Mini 1.0$0.568K$0.0448

The phone sends the same request a workstation would, so the row above is the bill. Base URL, key and id are set once in the playground or in whatever client the docs describe.

Small uncensored LLMs, 27B and under

Local rankings sort small models by the graphics card they fit: 8B, 12B, 24B, 27B at Q4. Over the API, size sets the price and the speed and nothing else, so the small builds are also the cheapest rows. Here are the builds of 27B or under that passed the Nov 18 check. None of them carries the stock refusal layer.

#ModelBuildOutput $/MCheck
1MythoMax L2 13BRoleplay$0.404Steady since Oct 20
2Cydonia 24B v4.3Roleplay$0.60Steady since Oct 23
3Gemma 4 12B Starry OceanRoleplay$0.875Steady since Oct 23
4Gemma 4 12B StationKeeperRoleplay$0.875Steady since Oct 23
5Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled DerestrictedAbliterated$1.0389Steady since Oct 23

Several of these have a UGI row under the same name, W/10 included, in the UGI table below.

The uncensored LLM field in 2026

In 2026 the catalog holds 170 builds of four kinds: 36 abliterated, 38 uncensored, 77 roleplay, 9 security. 138 passed the Nov 18 check. The GLM-5.3, DeepSeek V4 and Kimi K3 edits are the newest families here, and they displaced the Llama and Mistral tunes of 2025 from the top rows. The October 2026 list gives every build with the day it went on sale.

The catalog moves and the board moves with it. What you see is the 2026 ranking as of the date printed on it; the heading takes the new month on the first and the address stays put.

UGI scores next to the board

The UGI Leaderboard, run by DontPlanToEnd on Hugging Face, scores how uncensored a model is. UGI blends knowledge of sensitive subjects with W/10, willingness out of ten: how hard a model can be pushed before it declines or wanders off the instructions. The questions are private, and both open and proprietary models get a score.

What UGI cannot tell you is whether a build answers over an API today or what a million tokens cost. That is what this board adds. 9 catalog models match a UGI row by model name; highest UGI first:

ModelUGI rowUGIW/10Today’s checkOutput $/M
Anubis 70B v1.1Anubis-70B-v1.150.798.5Changed Nov 15$1.085
Cydonia 24B v4.3Cydonia-24B-v4.344.527.0Steady since Oct 23$0.60
Dolphin Mistral 24B Venice EditionDolphin-Mistral-24B-Venice-Edition40.797.8Steady since Oct 23$2.25
Wayfarer Large 70BWayfarer-Large-70B-Llama-3.337.065.0Changed Nov 14$2.4535
UnslopNemo 12B v4.1UnslopNemo-12B-v4.134.176.2Steady since Nov 14$1.972
Skyfall 36B v2Skyfall-36B-v233.906.0Steady since Oct 23$2.80
NeuralDaredevil 8B AbliteratedNeuralDaredevil-8B-abliterated33.206.5Changed Nov 14$1.76
Mag Mell R1 12BMN-12B-Mag-Mell-R131.146.0Changed Nov 15$1.7255
Stheno 8B v3.2L3-8B-Stheno-v3.220.563.8Changed Nov 14$0.7035

UGI and W/10 are copied from the leaderboard’s data file (its last change October 17, 2026; read October 25, 2026). UGI ran these weights between September 8, 2025 and December 16, 2025. Those scores are for the weights as UGI ran them. The check column is about the build this key reaches today.

The ranking rule, in full

Two public inputs. No composite score.

First input: the real-model check

Each day the gateway sends the same fixed prompt to every catalog model and stores the reply. Three verdicts are possible. Steady means the reply still matches that build’s earlier replies, and the date shows how long it has. Changed means the reply moved, dated. Not checked yet means the model joined the catalog after the run, or the run missed it.

A pass means the model that replied is the model you asked for, and it was up at the time. It does not grade willingness: an abliterated build and a stock one both pass if each answers as itself. UGI is the measure for that.

Second input: the live catalog

For every model the catalog records the id, build tags, input and output price per million tokens and context window. The price in a table here is the price your key pays. An empty context cell means the catalog holds no figure.

How rows are ordered

  1. A Steady verdict today puts a model on the board. The other two verdicts go in the list under it.
  2. Within the board, lower output price per million tokens ranks above higher. Output is the side that dominates an agent run or a report-drafting session.
  3. Equal output prices fall back to alphabetical order. Input price, context and build break no ties.

Price comes second because no other number on a row is simply a fact. Benchmarks depend on the prompt set, the sampler and whoever judges. The price is what your key is charged, to a fraction of a cent, and it moves only when the catalog does.

Nobody buys a row. No author-reported figures, no paid placement. A model climbs by passing the check and charging less; it drops by failing or charging more.

Neither data file needs a key. Two commands reproduce the board:

bash
1# 1. Models that answered today's check as the build they are named for
2curl -s https://pentestllm.ai/proof.json \
3 | jq -r '.data[] | select(.verdict == "stable") | .model' > stable.txt
4
5# 2. Their output prices and context windows, lowest output price first
6curl -s https://pentestllm.ai/pricing-index.json \
7 | jq -r --rawfile ok stable.txt '($ok | split("\n")) as $s
8 | .data[]
9 | select(.id | IN($s[]))
10 | [.id, .output, .context_window] | @tsv' \
11 | sort -t$'\t' -k2,2n

Every row that passed

138 models answered the Nov 18 check as the build their id names. Output price orders the table; any header re-sorts it. A name opens the model page with a first request ready to paste. Prices are per million tokens, the same as the price list.

Uncensored models and their prices per million tokens
Wire formats
GLM-5.3 Flash Uncensoredglm-5.3-flash-uncensoredUncensored$0.1053$0.3685—openai
MythoMax L2 13Bmythomax-l2-13bRoleplay$0.404$0.4044Kopenai
Manta Flash 1.0manta-flash-1.0Roleplay$0.07$0.5616Kopenai
Manta Mini 1.0manta-mini-1.0Roleplay$0.07$0.568Kopenai
Cydonia 24B v4.3cydonia-24b-v4.3Roleplay$0.484$0.60131Kopenai
GLM-5.3 Uncensoredglm-5.3-uncensoredUncensored$0.282$0.846—openai
Gemma 4 12B Starry Oceangemma-4-12b-starry-oceanRoleplay$0.175$0.875131Kopenai
Gemma 4 12B StationKeepergemma-4-12b-station-keeperRoleplay$0.175$0.875131Kopenai
Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestrictedqwen3.5-27b-claude-4.6-opus-distilled-derestrictedAbliterated$1.0389$1.0389262Kopenai
Gemma 4 31B Split-Untiedgemma-4-31b-split-untiedRoleplay$0.35$1.05262Kopenai
Veiled Calla 12Bveiled-calla-12bRoleplay$1.05$1.0533Kopenai
Gemma 4 31B Agares v1gemma-4-31b-agares-v1Roleplay$1.071$1.071262Kopenai
Gemma 4 31B Animus V14.1gemma-4-31b-animus-v14.1Roleplay$1.071$1.071262Kopenai
Gemma 4 31B Garnet V2gemma-4-31b-garnet-v2Roleplay$1.071$1.071262Kopenai
Gemma 4 31B Gembrain X Coregemma-4-31b-gembrain-x-coreRoleplay$1.071$1.071262Kopenai
Gemma 4 31B Isometry RPgemma-4-31b-isometry-rpRoleplay$1.071$1.071262Kopenai
Gemma 4 31B K1 v5gemma-4-31b-k1-v5Uncensored$1.071$1.071262Kopenai
Gemma 4 31B StyleTunegemma-4-31b-styletuneRoleplay$1.071$1.071262Kopenai
Qwen3.5 27B Blossom V6.4 Derestrictedqwen3.5-27b-blossom-v6.4-derestrictedAbliterated$1.071$1.071262Kopenai
Qwen3.5 27B BlueStar v3 Derestrictedqwen3.5-27b-bluestar-v3-derestrictedAbliterated$1.071$1.071262Kopenai
Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted Liteqwen3.5-27b-claude-4.6-opus-distilled-derestricted-liteAbliterated$1.071$1.071262Kopenai
Qwen3.5 27B Queen Derestrictedqwen3.5-27b-queen-derestrictedAbliterated$1.071$1.071262Kopenai
Gemma 4 31B Queengemma-4-31b-queenRoleplay$1.071$1.071262Kopenai
DeepSeek V4 Pro Qwen3.5 9B Uncensored Hereticdeepseek-v4-pro-qwen3.5-9b-uncensored-hereticAbliterated$1.12$1.1233Kopenai
Gemma 4 31B Gembrain Hereticgemma-4-31b-gembrain-hereticAbliterated$1.224$1.224262Kopenai
Gemma 4 31B SDFT Heretic RPgemma-4-31b-sdft-heretic-rpAbliterated$1.224$1.224262Kopenai
Gemma 4 31B Dark Gemistrygemma-4-31b-dark-gemistryRoleplay$1.224$1.224262Kopenai
Gemma 4 26B-A4B ChimeraXgemma-4-26b-a4b-chimeraxRoleplay$0.42$1.33262Kopenai
Gemma 4 26B-A4B Luminous Mirrorgemma-4-26b-a4b-luminous-mirrorRoleplay$0.42$1.33262Kopenai
Gemma 4 26B-A4B MeroMerogemma-4-26b-a4b-meromeroRoleplay$0.42$1.33262Kopenai
Gemma 4 26B-A4B Moonlight Duskgemma-4-26b-a4b-moonlight-duskRoleplay$0.42$1.33262Kopenai
Gemma 4 26B-A4B Musicagemma-4-26b-a4b-musicaRoleplay$0.42$1.33262Kopenai
Gemma 4 26B-A4B Opus Distillgemma-4-26b-a4b-opus-distillRoleplay$0.42$1.33262Kopenai
Gemma 4 26B-A4B Shadow Sirengemma-4-26b-a4b-shadow-sirenRoleplay$0.42$1.33262Kopenai
Gemma 4 26B A4B Cybersecuritygemma-4-26b-a4b-it-cybersecuritySecurity$0.4224$1.3376262Kopenai
Gemma 4 26B A4B ChimeraXgemma-4-26b-a4b-it-chimeraxRoleplay$0.48$1.48131Kopenai
Gemma 4 26B A4B DarkSoulgemma-4-26b-a4b-it-darksoulRoleplay$0.48$1.48131Kopenai
Gemma 4 26B A4B Luminousgemma-4-26b-a4b-it-luminousRoleplay$0.48$1.48131Kopenai
Gemma 4 26B A4B MeroMerogemma-4-26b-a4b-it-meromeroRoleplay$0.48$1.48131Kopenai
Gemma 4 26B A4B Moonlightgemma-4-26b-a4b-it-moonlightRoleplay$0.48$1.48131Kopenai
Gemma 4 26B A4B Musicagemma-4-26b-a4b-it-musicaRoleplay$0.48$1.48131Kopenai
Gemma 4 26B A4B OpusDistillgemma-4-26b-a4b-it-opusdistillRoleplay$0.48$1.48131Kopenai
Gemma 4 26B A4B ShadowSirengemma-4-26b-a4b-it-shadowsirenRoleplay$0.48$1.48131Kopenai
Gemma 4 26B-A4B ChimeraX Hereticgemma-4-26b-a4b-chimerax-hereticAbliterated$0.484$1.52262Kopenai
Gemma 4 26B-A4B DarkSoulgemma-4-26b-a4b-darksoulRoleplay$0.484$1.52262Kopenai
Artemis v1.1artemis-v1.1Roleplay$0.35$1.575262Kopenai
Gemma 4 31B Gembraingemma-4-31b-gembrainRoleplay$0.35$1.575262Kopenai
Gemma 4 31B Isometrygemma-4-31b-isometryRoleplay$0.35$1.575262Kopenai
Gemma 4 31B Scarletgemma-4-31b-scarletRoleplay$0.35$1.575262Kopenai
Gemma 4 31B Fabledgemma-4-31b-fabledRoleplay$0.35$1.575262Kopenai
Gemma 4 31B Gemsiclegemma-4-31b-gemsicleRoleplay$0.35$1.575262Kopenai
Gemma 4 31B MeroMero v2gemma-4-31b-meromero-v2Roleplay$0.35$1.575262Kopenai
Gemma 4 31B Novelistgemma-4-31b-novelistRoleplay$0.35$1.575262Kopenai
DeepSeek V4.1 Flash Abliterateddeepseek-v4.1-flash-abliteratedAbliterated$0.8488$1.6975—openai
GLM-5.3 Flash Cybersecurityglm-5.3-flash-cybersecuritySecurity$0.5093$1.69751049Kopenai
NemoMix Unleashed 12Bnemomix-unleashed-12bRoleplay$1.7255$1.725533Kopenai
Manta Pro 1.0manta-pro-1.0Roleplay$0.21$1.7566Kopenai
Qwen 2.5 32B Abliteratedqwen2.5-32b-abliteratedAbliterated$1.75$1.7533Kopenai
Gemma 4 31B GemBrain Hereticgemma-4-31b-it-gembrain-hereticAbliterated$0.40$1.76262Kopenai
Gemma 4 31B MeroMerogemma-4-31b-meromeroRoleplay$0.40$1.76262Kopenai
Gemma 4 31B DarkIdolgemma-4-31b-darkidolRoleplay$0.40$1.80262Kopenai
DeepSeek V4 Pro Uncensoreddeepseek-v4-pro-uncensoredUncensored$0.97$1.94—openai
GLM-5.3 Flash Abliteratedglm-5.3-flash-abliteratedAbliterated$0.4656$1.94—openai
Magnum Diamond 24Bmagnum-diamond-24bRoleplay$1.972$1.97233Kopenai
UnslopNemo 12B v4.1unslopnemo-12b-v4.1Roleplay$1.972$1.9728Kopenai
Cydonia 24B v4.1cydonia-24b-v4.1Roleplay$1.20$2.00131Kopenai
Qwen3.8 27B Antislopqwen3.8-27b-antislopRoleplay$0.35$2.10262Kopenai
Qwen3.8 27B Fimiqwen3.8-27b-fimiRoleplay$0.35$2.10262Kopenai
Qwen3.8 27B Kimikoqwen3.8-27b-kimikoRoleplay$0.35$2.10262Kopenai
Qwen3.8 27B Kiwenqwen3.8-27b-kiwenRoleplay$0.35$2.10262Kopenai
Qwen3.8 27B Novelistqwen3.8-27b-novelistRoleplay$0.35$2.10262Kopenai
Qwen3.8 27B Synthiaqwen3.8-27b-synthiaRoleplay$0.35$2.10262Kopenai
Dolphin Mistral 24B Venice Editiondolphin-mistral-24b-veniceUncensored$0.50$2.2533Kopenai
Apex Flash 1 Abliteratedapex-flash-1-abliteratedAbliterated$0.68$2.32262Kopenai
Uncensored AI Roleplayuncensored-ai-roleplayRoleplay$0.60$2.34128Kopenai
Qwen3.8 27B Cybersecurityqwen3.8-27b-cybersecuritySecurity$0.40$2.40262Kopenai
Qwen3.8 27B DarkIdolqwen3.8-27b-darkidolRoleplay$0.40$2.40262Kopenai
Jailbreak-Free Assistantjailbreak-free-assistantHouse build$0.75$2.501049Kopenai
DeepSeek R1 Distill Llama 70B Abliterateddeepseek-r1-distill-llama-70b-abliteratedAbliterated$2.80$2.8016Kopenai
Gemma 4 26B-A4B Uncensoredgemma-4-26b-a4b-uncensoredUncensored$0.60$2.8033Kopenai
Llama 3.3 70B Abliteratedllama-3.3-70b-abliteratedAbliterated$2.80$2.8033Kopenai
Skyfall 36B v2skyfall-36b-v2Roleplay$1.925$2.8033Kopenai
MiMo V2.6 Pro Uncensoredmimo-v2.6-pro-uncensoredUncensored$1.4768$2.95371049Kopenai
Uncensored AI Novelistuncensored-ai-novelistHouse build$0.75$3.00262Kopenai
Uncensored Coding Agentuncensored-coding-agentHouse build$0.90$3.001049Kopenai
DeepSeek V4.1 Flash 1017 Abliterateddeepseek-v4.1-flash-1017-abliteratedAbliterated$0.7275$3.03131000Kopenai
GLM-5.3 Flash Apex Abliteratedglm-5.3-flash-apex-abliteratedAbliterated$0.80$3.201049Kopenai
DeepSeek V4.1 Flash 1017 Uncensoreddeepseek-v4.1-flash-1017-uncensoredUncensored$0.8003$3.2011049Kopenai
DeepSeek V4.1 Flash 1017 Uncensored Thinkingdeepseek-v4.1-flash-1017-uncensored-thinkingUncensored$0.8003$3.2011049Kopenai
DeepSeek V4.1 Flash 1017 Hereticdeepseek-v4.1-flash-1017-heretic-uncensoredAbliterated$0.825$3.30512Kopenai
DeepSeek V4.1 Flash 1017 Heretic Thinkingdeepseek-v4.1-flash-1017-heretic-uncensored-thinkingAbliterated$0.825$3.30512Kopenai
DeepSeek V4.1 Flash 1017 Abliterated Fastdeepseek-v4.1-flash-1017-abliterated-fastAbliterated$1.75$3.501000Kopenai
DeepSeek V4.1 Flash Abliterated Fastdeepseek-v4.1-flash-abliterated-fastAbliterated$1.75$3.501000Kopenai
Qwen3.8 27B Uncensoredqwen3.8-27b-uncensoredUncensored$1.2125$3.637533Kopenai
MiMo V2.6 Flash Uncensored Thinkingmimo-v2.6-flash-uncensored-thinkingUncensored$1.25$3.751049Kopenai
ReMM SLERP L2 13Bremm-slerp-l2-13bRoleplay$2.036$3.786Kopenai
DeepSeek V4 Pro Abliterateddeepseek-v4-pro-abliteratedAbliterated$1.00$4.001000Kopenai
OrcaCyber MiniMax M3.1 Flashminimax-m3.1-flash-cybersecuritySecurity$2.00$4.00262Kopenai
OrcaCyber Qwen3.8 Flash Nextqwen3.8-flash-next-cybersecuritySecurity$2.00$4.00262Kopenai
GLM-5.3 Abliteratedglm-5.3-abliteratedAbliterated$2.1825$4.365—openai
Qwen3.8 27B Abliteratedqwen3.8-27b-abliteratedAbliterated$1.0037$4.5326262Kopenai
Abliterated Large Flashabliterated-model-large-flashAbliterated$2.50$5.001049Kopenai
GLM-4.6 Derestricted v5glm-4.6-derestricted-v5Abliterated$1.40$5.25131Kopenai
Qwen3.8 27B Fableqwen3.8-27b-fableRoleplay$0.875$5.25524Kopenai
Qwen3.8 27B Queenqwen3.8-27b-queenRoleplay$0.875$5.25524Kopenai
GLM-5.3 Flash Code Uncensoredglm-5.3-flash-code-uncensoredUncensored$6.00$6.00—openai
MiMo V2.6 Flash Abliteratedmimo-v2.6-flash-abliteratedAbliterated$2.00$6.001049Kopenai
MiMo V2.6 Flash Uncensoredmimo-v2.6-flash-uncensoredUncensored$2.00$6.001049Kopenai
Aion RP Llama 3.1 8Baion-rp-llama-3.1-8bRoleplay$3.20$6.4033Kopenai
Nemotron 3 Ultra 550B Uncensorednemotron-3-ultra-550b-a55b-uncensoredUncensored$1.0913$6.402262Kopenai
Qwen3.8 27B Obliteratedqwen3.8-27b-obliteratedAbliterated$0.582$7.27524Kopenai
Abliterated Modelabliterated-modelAbliterated$2.50$7.50262Kopenai
Qwen 3.8 27B Uncensored Fastqwen3.8-27b-uncensored-fastUncensored$2.50$7.50524Kopenai
Qwen 3.8 27B Uncensored Thinkingqwen3.8-27b-uncensored-thinkingUncensored$2.50$7.501049Kopenai
No Filter AIno-filter-aiHouse build$5.85$7.80197Kopenai
Uncensored AI Assistantuncensored-ai-assistantHouse build$5.85$7.80197Kopenai
Uncensored Chatbotuncensored-chatbotHouse build$5.85$7.80197Kopenai
Unfiltered AIunfiltered-aiHouse build$5.85$7.80197Kopenai
Gemma 31B Heretic Uncensoredgemma-31b-heretic-uncensoredAbliterated$1.50$8.25—openai
DansPE 1.3 24Bdans-pe-1.3-24bRoleplay$2.908$11.63233Kopenai
Magnum v4 72Bmagnum-v4-72bRoleplay$8.0243$11.968416Kopenai
Inkling Uncensoredinkling-uncensoredUncensored$2.13$12.1403262Kopenai
Abliterated Model Large V2abliterated-model-large-v2Abliterated$7.4996$12.4997960Kopenai
Abliterated Model Largeabliterated-model-largeAbliterated$7.50$12.501000Kopenai
Uncensored AI Chatuncensored-ai-chatHouse build$11.70$15.60197Kopenai
Uncensored Code Assistantuncensored-code-assistantHouse build$11.70$15.60131Kopenai
Unrestricted AIunrestricted-aiHouse build$11.70$15.60131Kopenai
Kimi K3 Uncensoredkimi-k3-uncensoredUncensored$5.745$28.68262Kopenai
Magistral Small Uncensoredmagistral-small-uncensoredUncensored$0.005 a requestflat—openai
Ministral 3 14B Instruct Uncensoredministral-3-14b-instruct-uncensoredUncensored$0.005 a requestflat—openai
Ministral 3 8B Instruct Uncensoredministral-3-8b-instruct-uncensoredUncensored$0.005 a requestflat120Kopenai
Mistral Large 2 Uncensoredmistral-large-2-uncensoredUncensored$0.005 a requestflat120Kopenai
Mistral Large 3 Uncensoredmistral-large-3-uncensoredUncensored$0.09 a requestflat120Kopenai
Qwen3 Coder Next Uncensoredqwen3-coder-next-uncensoredUncensored$0.005 a requestflat120Kopenai
Qwen3 Next 80B-A3B Instruct Uncensoredqwen3-next-80b-a3b-instruct-uncensoredUncensored$0.005 a requestflat120Kopenai
Qwen3 VL 235B-A22B Instruct Uncensoredqwen3-vl-235b-a22b-instruct-uncensoredUncensored$0.005 a requestflat120Kopenai
Qwen3.5 397B-A17B Uncensoredqwen3.5-397b-a17b-uncensoredUncensored$0.09 a requestflat120Kopenai
Qwen3.8 2.4T-A95B Uncensoredqwen3.8-2.4t-a95b-uncensoredUncensored$0.07 a requestflat120Kopenai

Prices per million tokens, read from the live catalog on 2026-11-18.

Off the board today

32 models sit off the board: 25 replied differently from before, and 7 were not covered by this run. All of them still answer requests. No verdict is not a failing verdict.

ModelBuildOutput $/MCheck
Deepseek V4.1 Flash UncensoredUncensored$0.1882Changed Nov 18
Magidonia 24B v4.3Roleplay$0.4235Changed Nov 14
Cydonia 24B v2Roleplay$0.4828Changed Nov 14
Stheno 8B v3.2Roleplay$0.7035Changed Nov 14
Anubis 70B v1.1Roleplay$1.085Changed Nov 15
Anubis 70B v1Roleplay$1.24Changed Nov 12
Euryale 70B v2.2Roleplay$1.428Changed Nov 14
Cu-Mai R1 70BRoleplay$1.7255Changed Nov 14
Hanami 70B x1Roleplay$1.7255Changed Nov 15
Mag Mell R1 12BRoleplay$1.7255Changed Nov 15
Starcannon Unleashed 12B v1.0Roleplay$1.7255Changed Nov 14
NeuralDaredevil 8B AbliteratedAbliterated$1.76Changed Nov 14
Euryale 70B v2.3Roleplay$1.972Changed Nov 15
The Omega Directive 24B v2.0Roleplay$2.00Changed Nov 14
GPT-6 Sol UncensoredUncensored$2.25Changed Nov 18
Wayfarer Large 70BRoleplay$2.4535Changed Nov 14
Llama 3 70B Abliterated v3.5Abliterated$2.80Changed Nov 12
Qwen3.8 Max UncensoredUncensored$3.00Changed Nov 15
DeepSeek V4.1 Flash 0910 UncensoredUncensored$3.30Changed Nov 18
DeepSeek V4.1 Flash 0910 Uncensored ThinkingUncensored$3.30Changed Nov 18
DeepSeek V4.1 Flash Uncensored ThinkingUncensored$3.30Changed Nov 18
DeepSeek R1 Distill Qwen 32B AbliteratedAbliterated$5.60Changed Nov 17
Kimi K3.1 Pro UncensoredUncensored$5.82Changed Nov 14
OrcaCyber Zero 1.5Security$18.75Changed Nov 14
Aion 3.5 Mini UncensoredUncensoredbilled per requestChanged Nov 11
Gemma 4 31B Mero Artemis v0.3.1Roleplay$1.12Not checked yet
Gemini 4 Argon UncensoredUncensored$2.25Not checked yet
Rocinante 12B v1.1Roleplay$2.38Not checked yet
Red Team (Flash)Security$3.00Not checked yet
Red Team (Pro)Security$6.00Not checked yet
Kimi K2.6 UncensoredUncensored$7.34Not checked yet
Cyber FrontierSecurity$15.75Not checked yet

The same rows, minus the check column, are on the models page.

What a long session costs as the context grows

An agent resends its scope notes, tool output and history with every turn, so the input side swells over a session while each reply stays about the same length. The table prices 100 replies of 300 tokens at three context sizes for the first 12 priced rows on today’s board, from the same catalog read.

Model4,000 in8,000 in16,000 in
GLM-5.3 Flash Uncensored$0.0532$0.0953$0.1795
MythoMax L2 13B$0.1737$0.3353$0.6585
Manta Flash 1.0$0.0448$0.0728$0.1288
Manta Mini 1.0$0.0448$0.0728$0.1288
Cydonia 24B v4.3$0.2116$0.4052$0.7924
GLM-5.3 Uncensored$0.1382$0.251$0.4766
Gemma 4 12B Starry Ocean$0.0963$0.1663$0.3063
Gemma 4 12B StationKeeper$0.0963$0.1663$0.3063
Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted$0.4467$0.8623$1.6934
Gemma 4 31B Split-Untied$0.1715$0.3115$0.5915
Veiled Calla 12B$0.4515$0.8715$1.7115
Gemma 4 31B Agares v1$0.4605$0.8889$1.7457

USD per 100 replies at 300 output tokens each, board order, catalog of 2026-11-18.

With 16,000 tokens in, 100 replies cost $0.1795 on GLM-5.3 Flash Uncensored and $1.7457 on Gemma 4 31B Agares v1. Every turn the history adds widens that gap, which is why trimming tool output before it goes back into context saves more than switching models.

API or local weights

Local rankings measure a model by what fits on the card: at Q4, roughly 5 GB for an 8B, 16 GB for a 27B, and a 70B wants 40 GB or more. Run the weights yourself and there is no per-token bill and no prompt leaves the box, which matters when the engagement contract says client data stays on your hardware. The cost is the GPU and the hours spent downloading, updating and swapping models.

Run the weightsCall the API
GPUYours, with 20 GB or more if a 20B at Q4 is to fitNot needed
SetupLM Studio or Ollama, then the client that talks to itThree fields in the client you already use: base URL, key, model id
PriceHardware and electricityFrom $0.3685 per million output tokens
Model choiceWhatever the card holds170 builds, one id apart
Proof it is the model namedThe file on diskA dated daily check per model

Go local when the contract keeps prompts on your own machines or a GPU is already sitting idle. Go API when there is no GPU, when you want to try several builds by changing one string, or when the same model has to serve a CI job, a laptop and a phone at the same time. The Ollama guide covers the local route.

The data files, and how to cite them

Both files refresh every five minutes and permit cross-origin reads, so a script, a notebook or a spreadsheet can pull them straight in. To cite the board, quote the check time with the month:

PentestLLM, "Best uncensored LLMs by API, November 2026 board", https://pentestllm.ai/blog/uncensored-llm-leaderboard, check of 2026-11-18 00:53 UTC.

Send a request to a board model

Send the id from the model column. The request is the standard chat completions shape, so every OpenAI-compatible client works once the base URL and key are swapped.

curl
1curl https://api.pentestllm.ai/v1/chat/completions \
2 -H "Authorization: Bearer sk-ptl-live-…" \
3 -H "Content-Type: application/json" \
4 -d '{"model": "glm-5.3-flash-uncensored", "messages": [{"role": "user", "content": "Summarize what this nmap output says about the host in scope: 22/tcp open ssh OpenSSH 8.9, 80/tcp open http nginx 1.18, 8443/tcp open ssl/https-alt"}]}'

The base URL is https://api.pentestllm.ai/v1. Settings for OpenCode, Claude Code, Cursor, Cline and aider are on the docs page and the agent pages.

A bill worked from the same catalog read: 1,000 turns, each 2,000 input tokens and 300 output tokens, about what an agent spends on a source tree review with the scope document in context. That comes to $0.32 on GLM-5.3 Flash Uncensored and $20.09 on Kimi K3 Uncensored, the dearest row that passed today. Credit is prepaid, and each request settles at these rates as it completes.

Look at the board again before a long run. The verdict column moves once a day; the price column follows the catalog.

Questions

Two do. DontPlanToEnd's UGI Leaderboard on Hugging Face scores open and proprietary models on sensitive-topic knowledge and willingness to answer. This page covers the models you can call over an API: passed or not on today's real-model check, then output price, with UGI's score beside the 9 models it has run under the same name. See the daily check.

UGI is short for Uncensored General Intelligence. It is a Hugging Face Space that scores language models on what they know about sensitive topics (hazardous, entertainment and socio-political) blended with W/10, a willingness measure. It also publishes NatInt, a general knowledge and reasoning figure, and does not release its questions.

Open the Space and sort on UGI or W/10. The model-type filters (foundation, fine-tuned, merged, proprietary) cut the list down. To call a model from that list without downloading it, find the same name on the models page here and put its id in a chat completions request. See the models page.

Willingness out of ten: how far a model goes before it refuses or stops following the instructions. It feeds the UGI score and has two parts, W/10-Direct for flat refusals and W/10-Adherence for drifting off the instructions.

For November 2026, GLM-5.3 Flash Uncensored holds the first row: among the models that passed the Nov 18 check, it has the lowest output price at $0.3685 per million tokens. Among passed rows with a UGI score, Cydonia 24B v4.3 is highest (UGI 44.52). The task table on this page names a row per job.

For November 2026, GLM-5.3 Flash Uncensored holds the first row: among the models that passed the Nov 18 check, it has the lowest output price at $0.3685 per million tokens. Among passed rows with a UGI score, Cydonia 24B v4.3 is highest (UGI 44.52). The task table on this page names a row per job. The board has no quality score of its own. It tells you which build answered as itself today, what a million output tokens cost, and UGI's figure where UGI has one. See the models by family.

On price, today's first row: GLM-5.3 Flash Uncensored at $0.3685 per million output tokens, reached over the API from a terminal, an agent or a phone browser with nothing installed. The task table names the first security, dialogue and abliterated rows the same way. See the playground.

MythoMax L2 13B, the cheapest dialogue tune to pass the Nov 18 check, at $0.404 per million output tokens: $0.1737 for 100 replies on 4,000 tokens of context. It answers at once with no reasoning pass, and the playground runs it in a phone browser. See the playground.

MythoMax L2 13B: the cheapest build of 27B or under to pass the Nov 18 check, at $0.404 per million output tokens. Small abliterated and fine-tuned builds carry no stock refusal layer, and over the API their size sets the price rather than the hardware you need.

From today's board, MythoMax L2 13B: the cheapest dialogue tune to pass the Nov 18 check, at $0.404 per million output tokens. A dialogue tune writes at once with no reasoning pass, so short questions come back short; the DeepSeek and GLM-5.3 builds think first and want a reply limit of 1,000 tokens or more. See the models page.

GLM-5.3 Flash Uncensored: $0.3685 per million output tokens on the catalog of 2026-11-18, so $0.0532 for 100 replies of 300 tokens on 4,000 tokens of context. Input is charged separately at the build's own rate.

Heretic is an open-source tool that automates abliteration: it finds the refusal direction in a model and removes it from the weights, so a Heretic build is an abliterated build produced by that tool. 8 catalog builds carry Heretic in the id, 8 of them on today's board; the catalog tags all of them abliterated. See abliterated vs uncensored.

Yes. Dolphin is an uncensored fine-tune line. Dolphin Mistral 24B Venice Edition is on the catalog as dolphin-mistral-24b-venice at $2.25 per million output tokens, steady since oct 23 on the daily check, and UGI scores it under the same name.

No. Every board model answers through the API: point a client at https://api.pentestllm.ai/v1, paste a key and send the model id from a terminal, an agent or a phone. Credit is prepaid from $10 and charged per token, and nothing is downloaded. See the client setup docs.

Yes. Artificial Analysis runs an Image Arena leaderboard as a Hugging Face Space (read October 25, 2026). The image and video models callable here are on the media page with per-image and per-second prices. See image and video models.

Every model above is on the uncensored model list with its live price per million tokens. Per-token pricing sets the whole catalog side by side, and the family directory groups the builds by base model.

Get a keyPrepaid from $10, no subscription.

New models and price changes, once a month.

New uncensored builds with their prices, price moves, and what the daily model check found.

Help