GPT‑6 released early? Mythos reportedly stalled as staggering compute demands ‘burned out’ Anthropic

Rumors and a stalled roadmap

It has been reported that Anthropic’s next-generation model, codenamed Mythos, has stalled in development amid unexpectedly huge compute requirements. Chinese technology site Ifeng (凤凰网) published the account, and the piece also rekindled online rumors that a GPT‑6 rollout might have occurred early — claims that remain unverified. So which is true? For now, industry watchers say the noise outstrips the facts.

Staggering compute and frayed engineering teams

Reportedly, the project’s resource needs ballooned far beyond initial estimates, forcing Anthropic to re‑evaluate training schedules and infrastructure investments. Training at the new scale demands thousands of accelerator‑years and complex orchestration; that means both enormous cost and intense operational stress. The Ifeng report described engineering teams as “burned out,” a shorthand for the human and technical limits AI labs confront when chasing frontier models.

Geopolitics, chips and competition

This episode comes as the global market for advanced accelerators — led by NVIDIA’s H100 class chips — is tight and politicized. Export controls and trade policy have reshaped supply chains, complicating choices for US and Chinese firms alike. Chinese competitors such as Baidu (百度) and others are pressing their own development timelines, so delays at one lab reverberate through a highly competitive international race for capability and customers.

Uncertainty ahead

Anthropic has not publicly confirmed the stall or any early GPT‑6 release; it has been reported that details remain opaque and subject to change. If the report proves accurate, the pause would underline a broader reality: next‑generation models are as much an organizational and supply‑chain challenge as they are a research milestone. Will the field slow to catch its breath — or will rivals step into the opening? The next few weeks should tell.