AnonRouter
Privacy-focused, OpenAI-compatible router with anonymous accounts and an open-source relay running in an attested Intel TDX enclave. Its uncensored models come from Venice, abliteration.ai and Phala AI, each labeled Anonymous, Private, TEE or E2EE.
BILLINGPrepaid USD credits, $5 to $1,000 per purchase in whole dollars, valid for 365 days. A 5% service fee ($0.75 minimum) is added to each credit purchase as a launch promotion (regular rate 8%, $1 minimum). Per-token prices equal the upstream provider's rate with no per-model markup, so Venice-served models cost exactly Venice's rates. Card checkout, or Bitcoin on-chain. Free models (none uncensored) are opt-in and run on Anonymous routes.
| Model | Context | In $/1M | Out $/1M | Per 1K chats |
|---|---|---|---|---|
GLM 4.7 Flash HereticGLM-4.7-Flash (Olafangensan/GLM-4.7-Flash-heretic) | 200K | $0.070 | $0.40 | $0.27 |
Gemma 4 UncensoredGemma 4 26B (Jiunsong/supergemma4-26b-uncensored-gguf-v2) · 26B | 256K | $0.16 | $0.50 | $0.41 |
Gemma 4 26B A4B UncensoredGemma 4 26B A4B (coder3101/gemma-4-26B-A4B-it-heretic) · 25.2B | 64K | $0.19 | $0.88 | $0.63 |
Mistral Uncensored 1.2Dolphin Mistral 24B Venice Edition (Mistral Small 24B) · 24B | 128K | $0.20 | $0.90 | $0.65 |
DeepSeek V4.1 Flash UncensoredDeepSeek V4.1 Flash (dealignai/DeepSeek-V4.1-Flash-UNCENSORED-FP8) · 552B | 1M | $0.33 | $1.32 | $0.99 |
Qwen 3.8 27B UncensoredQwen3.8-27B (HauhauCS Uncensored Aggressive) · 27B | 262K | $0.30 | $1.50 | $1.05 |
Mistral Uncensored RoleplayMistral Small 24B (dphnAI/24B-3.2-RP-K2-final) · 24B | 128K | $0.50 | $2.00 | $1.50 |
Qwen 3.6 Plus UncensoredQwen 3.6 Plus (Alibaba) | 1M | $0.63 | $3.75 | $2.50 |
GLM 5.3 UncensoredGLM-5.3 (abliteration.ai abliterated-model-large-v2) | 1M | $3.00 | $5.00 | $5.50 |
- Type
- Aggregator
- Method
- Mixed
- Access
- Web chat + OpenAI-compatible API
- Logging
- No logs
- Trains on prompts
- No
- Guarantee
- Policy only
- Jurisdiction
- Wyoming, US
- Signup
- No personal info
- Payments
- Card, Crypto
- Data checked
- Oct 28, 2026
PRIVACYAnonRouter doesn't store prompts or responses, only per-request metadata (model, tokens, cost, status). Its relay runs in an Intel TDX confidential VM that anyone can attest, with open-source code. In the default ticket flow the part that sees your prompt never learns your account. The OpenAI-compatible key mode, on by default, joins the two in memory. The model host still reads your prompt on most uncensored routes. Seven of the nine are Private (the provider promises no retention) or Anonymous (no retention promise). Gemma 4 26B A4B Uncensored has an E2EE route via Venice, where a verified Venice gateway may decrypt and forward. Qwen 3.8 27B Uncensored runs on a Phala route labeled TEE, which AnonRouter's docs elsewhere call not confidential. Each model page and each response header names the serving provider. Policy ↗
Operated by Sunday Labs LLC (Sheridan, Wyoming). The terms set Wyoming law and binding AAA arbitration, with a 30-day opt-out, and require users to be 18 or older. A general router with about 370 models. Its catalog has a moderation filter, and an anonrouter/uncensored router model picks among the uncensored routes. Upstreams for the models listed here: Venice serves seven at Venice's own prices; abliteration.ai serves GLM 5.3 Uncensored; Phala AI serves Qwen 3.8 27B Uncensored. Cheaper anonymous fallback routes from Surplus Intelligence and Cheaper Inference sit 15-35% below Venice. Gemma 4 26B A4B Uncensored also has a Phala TEE route at $0.15/$0.70 with 262K context, which carries its plaintext traffic. Requests can pin or exclude providers, set a minimum privacy level, and cap price. Fallback, up to 3 providers, never drops below the allowed privacy level. DeepSeek V4.1 Flash 1017 Uncensored and abliteration.ai's Abliterated Model are listed but not callable. The relay code is public at github.com/anonrouter/confidential-content-plane (Apache-2.0); no independent audit is published.
- https://anonrouter.ai/models?moderation=uncensored
- https://control.anonrouter.ai/v1/catalog/models
- https://control.anonrouter.ai/v1/billing/catalog
- https://anonrouter.ai/models/z-ai/glm-4.7-flash-heretic
- https://anonrouter.ai/pricing
- https://anonrouter.ai/privacy-policy
- https://anonrouter.ai/terms
- https://anonrouter.ai/privacy
- https://anonrouter.ai/security
- https://anonrouter.ai/sign-up
- https://docs.anonrouter.ai/llms-full.txt
- https://docs.anonrouter.ai/api/markdown/models
- https://docs.anonrouter.ai/api/markdown/privacy
- https://docs.anonrouter.ai/api/markdown/provider-routing
- https://docs.anonrouter.ai/api/markdown/routing
- https://docs.anonrouter.ai/api/markdown/tee-verification
- https://docs.anonrouter.ai/api/markdown/credits
- https://github.com/anonrouter/confidential-content-plane
- https://github.com/anonrouter/anonrouter-sdk
- https://api.venice.ai/api/v1/models?type=text
- https://docs.venice.ai/overview/pricing
More on AnonRouter
- AnonRouter review: Venice's uncensored models at Venice's prices, an attested relay, and privacyAnonRouter routes 9 uncensored models, 7 of them from Venice at Venice's own prices, with anonymous signup and an open-source enclave relay. It tops our privacy ranking at 80/100. What the score covers and what it doesn't.
- Is the uncensored model you're paying for the one you get?Often it's resold. Venice's Dolphin model costs $0.20 / $0.90 at Venice, twice that at NanoGPT, and 14x the input price at DemonRoute. What to check before you pay.
- Which uncensored AI is best for coding and security research?Unmeasured so far. Three options target it: Abliteration.ai (red-teaming, 1M context, $3 / $5 per 1M tokens), NanoGPT's three Cybersecurity fine-tunes (about $0.27 to $0.40 per 1,000 chats), and RedModel's free research preview.
- NoTrack review: one undisclosed uncensored model, no signup, and a contradictory privacy storyFree anonymous web chat plus an API for one uncensored model at $0.15 / $0.50 per 1M tokens. No signup, but its own pages disagree on whether prompts are kept.
- NoFilterGPT review: a $11.99 uncensored chat with an unnamed model and operatorOne custom model, a free tier of 7 messages a day, and a $11.99/mo plan with 750 API messages a day. No company is named, and the model's origin isn't disclosed.
- Can you pay for uncensored AI with crypto?Yes: 15 of 35 providers we index take crypto, and 4 take nothing else. PayPerQ accepts Lightning from $0.10 with no account, PayPerQ and Unjailed take Monero, and AnonRouter takes Bitcoin with no email.
- Is it legal to use uncensored AI?Generally yes: the model is software, often under a permissive license. What you generate is still covered by the law, and services set their own age limits, from 13 with consent to 18+.
- Can you use uncensored AI without an account?Yes. 10 of the 35 providers we index need no account, including NanoGPT and PayPerQ from $0.10 in crypto. But no account doesn't mean no logs, and lost keys often can't be recovered.