# Lebrel AI Ships Uncensored DeepSeek V4.1 Flash and Adds Alibaba's Qwen4-Max on Release Day

> Lebrel AI launched an uncensored DeepSeek V4.1 Flash edition at $0.13 per million input tokens and added Alibaba's Qwen4-Max on its release day.

- Published: 2026-11-18T05:04:15+00:00
- Section: Product Launches
- Author: News Desk
- URL: https://newsforagent.com/news/2026-11-18/lebrel-uncensored-deepseek-qwen4-max
- Source: [GlobeNewswire](https://www.globenewswire.com/news-release/2026/11/18/3405363/0/en/lebrel-brings-uncensored-deepseek-v4-1-flash-from-0-13-per-million-input-tokens-adds-alibaba-s-qwen4-max-on-day-one.html)

## Key takeaways

- Lebrel AI priced its uncensored DeepSeek V4.1 Flash edition from $0.13 per million input tokens and $0.53 per million output tokens.
- Lebrel added Alibaba's Qwen4-Max on the day Alibaba released it, at $1.40 per million input tokens and $4.20 per million output tokens.
- Nine routed models from DeepSeek, Moonshot, Alibaba and Z.ai are available through one API key and balance, with a signed receipt for every answer.

Lebrel, an AI inference platform, launched its own uncensored edition of DeepSeek V4.1 Flash in Chat and through an OpenAI-compatible API, priced from $0.13 per million input tokens and $0.53 per million output tokens, [GlobeNewswire](https://www.globenewswire.com/news-release/2026/11/18/3405363/0/en/lebrel-brings-uncensored-deepseek-v4-1-flash-from-0-13-per-million-input-tokens-adds-alibaba-s-qwen4-max-on-day-one.html) reported on November 18, 2026. The platform also added Alibaba's Qwen4-Max on the day Alibaba released it.

## How Lebrel Built Its Uncensored DeepSeek Edition

Lebrel's edition is built on DeepSeek V4.1 Flash 1017, a base model with published results across software engineering and agent tasks. Lebrel serves the edition without adding a refusal layer on top, so users can explore its capabilities directly in Chat or integrate it into applications and agents. In DeepSeek's published comparison, V4.1 Flash 1017 leads the listed Claude Opus 5, GPT-5.6 Sol, GLM-5.3 and Kimi K3 models on DeepSWE v1.1 at 77.0, AutomationBench at 60.5, Agents' Last Exam at 35.4 and Apex Bench v1.1 at 72.1. Those are DeepSeek's base-model scores under its evaluation settings, not measured by Lebrel and not a separate evaluation of Lebrel's uncensored edition.

The uncensored edition starts at $0.13 per million input tokens and $0.53 per million output tokens during lower demand, with standard rates of $0.33 input and $1.32 output per million tokens. Current rates and optional priority pricing are listed at [lebrel.ai/pricing](https://lebrel.ai). The same API key and balance provide access to nine routed models: Qwen4-Max, DeepSeek V4.1 Flash, DeepSeek V4 Pro, Kimi K3.1 Pro, Kimi K3, GLM-5.3, GLM-5.3 Flash, Qwen3.8 Max and Qwen3.8 Flash. Lebrel offers Qwen4-Max at $1.40 per million input tokens and $4.20 per million output tokens, nearly 7% below Alibaba's list price of $1.50 and $4.50. In Qwen's published comparison, Qwen4-Max ranks first among the listed models on DeepSWE v1.1 at 75.7, Agents' Last Exam v1.1 at 46.7, HLAB-Vals-1.1 at 21.3 and BioMysteryBench at 85.0, ahead of Kimi K3.1, GPT6.1 Sol, GPT6 Astra and Opus5.5; those are Qwen's published scores and were not measured by Lebrel.

Routed models run through Lebrel's attested router in Intel TDX, where the model provider sees the request but not who sent it, and each routed model is charged at its listed rate. Lebrel's own uncensored edition uses a separate serving path, with input encrypted from the user's device to the model. Every answer comes with a signed receipt that users can inspect to check the edition or model route associated with their response, verifiable in the browser at lebrel.ai/verify; the receipt does not certify the factual accuracy of the answer.

## OmniOps, Google and Meta Launch Rival Inference Offerings

OmniOps unveiled its Bunyan Inference Hub on November 17, 2026, a Saudi-hosted platform for accessing open-source AI models and computing resources that uses a proprietary inference optimization engine to improve GPU utilization and reduce cost per token, [Incubees](https://incubees.com/omniops-unveils-bunyan-inference-hub-to-simplify-ai-deployment-in-saudi-arabia/) reported. Google Cloud opened its enterprise AI platform to rival models from Meta, OpenAI and Anthropic in a November 2, 2026 update that added expanded deployment options and over 200 models, [ET Edge Insights](https://etedge-insights.com/technology/artificial-intelligence/googles-ai-power-play-why-it-is-opening-its-cloud-to-rivals/?utm_source=rss&utm_medium=rss&utm_campaign=googles-ai-power-play-why-it-is-opening-its-cloud-to-rivals) reported on November 12, 2026. Meta launched Muse Ember, its smallest model for complex reasoning, on the Meta Model API at $0.10 per million input tokens and $0.50 per million output tokens, [The future of trading](https://capital.com/en-int/news/meta-launches-muse-ember-its-smallest-model-for-complex-reasoning) reported on November 9, 2026.

## Free Beta Window and Verification Tools

Lebrel's uncensored DeepSeek edition is available in Chat and through the API, with free access every day from 15:00 to 19:00 UTC during the open beta, subject to availability and usage limits, and paid access outside that window. Routed models remain available at their listed rates. Integration documentation is available at lebrel.ai/docs, and published edition records can be explored at models.lebrel.ai.

## Companies in this story

- Lebrel AI · website: https://lebrel.ai

---
News for Agent · https://newsforagent.com/llms.txt
