Epoch AI: Open-Weight GLM 5.3 Exploits Firefox Patch Gap at About $1,000 per Version
Epoch AI reports that GLM 5.3, an open-weight model released by Z.ai in August 2026, can exploit the browser patch gap to compromise six recent Mozilla Firefox releases. The marginal cost was about $1,000 per version, according to Epoch.
Epoch AI reports that GLM 5.3, an open-weight model from Z.ai, can exploit the browser patch gap to compromise six recent Mozilla Firefox releases at a marginal cost of about $1,000 per version. The research was published on October 29, 2026, and was covered by Unite.AI.
What Epoch reports
The patch gap is the window between a security fix becoming public and that fix reaching users. According to Epoch, GLM 5.3 used this window to compromise the six Firefox releases.
Epoch said the model operated autonomously. It also said the model repeatedly rediscovered an unpublished vulnerability during the attacks. The researchers disclosed that vulnerability to Mozilla.
Model details
GLM 5.3 is an open-weight model that Z.ai released in August 2026. Context window, parameter count and pricing are not covered in the Unite.AI summary of the research.
What this means
Epoch's finding concerns an open-weight model, which means the weights are publicly available, performing the exploitation autonomously. The reported marginal cost of about $1,000 per Firefox version is the figure Epoch gives for the attacks. The repeated rediscovery of an unpublished vulnerability, which Epoch disclosed to Mozilla, shows the model's activity was not limited to flaws that had already been patched publicly.
Related Articles
Gemini 4 Argon scores 97.5% on ARC-AGI-3 with vendor features, but only 49.8% in the standard harness
Google DeepMind's Gemini 4 Argon scored 97.5% on ARC-AGI-3 at a cost of $4,800 using the Provider Adapter Harness, according to ARC Prize. Under the Standard Harness it falls to 49.8%, and Epoch AI places it eighth on its Math ECI and sixth on its SWE ECI.
Artificial Analysis Index v5 drops Kimi K3.1 Pro from 4th to 9th; Gemini 4 Carbon takes #1 at 59
Artificial Analysis released Intelligence Index v5 on Tuesday, with 60% of the test built on private, held-out questions. Moonshot AI's Kimi K3.1 Pro fell from 4th to 9th, and Google's Gemini 4 Carbon debuted at #1 with a score of 59.
Replit's ViBench 2.0: Top models pass every test plan in only 49-65% of full software builds
Replit released ViBench 2.0, a benchmark in which an agent builds a full software product feature by feature from an empty project. GPT-6.1 Sol leads with a 65% pass@1, followed by GPT-6 Astra at 60% and Opus 5.5 at 49%. No open-weight model passes a full build.
Artificial Analysis launches Intelligence Index v5; Claude Opus 5.5 (max) leads at 58
Artificial Analysis has launched Intelligence Index v5, featuring 2 new benchmarks and 3 upgrades, including new coding and science benchmarks. For the first time, the majority of the Index is built on private datasets. Claude Opus 5.5 (max) leads at 58.
Comments
Loading...