PewDiePie Released Uncensored Ajax AI Model
The creator unveiled a fine-tuned version of the Qwen3.5-9B model designed for local execution.
Updated on Oct. 2, 2026 in Artificial Intelligence
Live Poll
Do you believe AI models should operate without developer-imposed censorship?
PewDiePie has unveiled the Ajax AI model, a fine-tuned iteration of the Qwen3.5-9B architecture, for use on home PCs. The release follows two account bans from OpenAI related to model distillation activities.
Why it matters
This development represents a growing push for self-hosted, unrestricted alternatives to large frontier models. By integrating Ajax into the Odysseus AI workspace, the creator aims to bypass the standard refusal behaviors enforced by centralized AI providers.
The Ajax model is based on the 9B parameter Qwen architecture. It leverages automatic abliteration, a technique that modifies model weights to strip away refusal behaviors.
The players
PewDiePie
A content creator who has expanded into AI development, focusing on local, unrestricted model alternatives.
OpenAI
The developer of frontier AI models that enforces strict usage policies and restricts distillation activities.
The details
The Ajax model is integrated into the self-hosted Odysseus AI workspace, which launched earlier in 2026. The creator utilized the open-source tool Heretic to perform model abliteration, a process that identifies and suppresses the internal logic responsible for model refusals. This approach relies on distillation—a training method where a smaller model learns from the outputs of a larger, more capable one—to maintain performance while stripping away restricted behaviors.
Timeline
May 2026: The Odysseus AI workspace was launched.
October 2, 2026: PewDiePie unveiled the Ajax AI model.
The Tech Race
The release of Ajax follows a pattern of developers modifying foundation models like Qwen3.5-9B to bypass centralized safety guardrails. It positions the Odysseus workspace as a competitor to commercial, cloud-based interfaces that maintain strict refusal behaviors.
Users can run the Ajax model locally on home PCs via the Odysseus AI workspace. This tool offers an alternative for those seeking an unrestricted AI experience without the dependency on cloud-based API access.
The takeaway
The rise of abliterated, self-hosted models signals a shift in how enthusiasts engage with AI, prioritizing model autonomy over centralized oversight. Interested users should monitor the creator's upcoming release for the refined iteration of Ajax.
What happens next
The creator has announced plans to conduct two more days of reinforcement learning for the model, with a subsequent version of both the Ajax model and the Odysseus workspace expected in the future.
Further reading
For more developments on local model deployment, see the latest in Artificial Intelligence.
Source note: This article includes information reported by Tom's Hardware.
Live Poll
Do you believe AI models should operate without developer-imposed censorship?
