Radar / AI security / Cantina Apex Flash-1
Cantina Apex Flash-1: open-weight security investigation model
Cantina (co-founded by Spearbit's Harikrishnan Mulackal) released Apex Flash-1 on Oct 1, 2026: an open-weight 321B model fine-tuned from GLM-5.3-Flash on 150 tasks derived from 50 real vulnerability cases, under MIT license.
Why it matters
Turns a security firm's real vulnerability research into a downloadable worker model for agentic security investigations, scoring 66.7% pass-at-one on 60 held-out tasks at an estimated $2.38 run cost; an experimental abliterated (low-refusal) variant was released alongside.
What you could build with it
A security consultancy pre-investigates incoming bug-bounty submissions with Apex Flash-1 as an agent worker model, automatically reproducing exploit chains against sandboxed targets so human researchers review pre-triaged reports instead of raw submissions.
Does it hold up?
Too early to judge: Cantina's own 60-task evaluation (40/60 passed vs 36/60 for base GLM-5.3-Flash) is company-reported and not independently reproduced; the abliterated variant raises explicit dual-use questions.
Learn more
RuntimeWire: Cantina's open-weight security model and the abliterated variant →
First spotted on article: source.
More AI security
Cloudflare security-audit-skill: multi-phase security audits for coding agentsA coding-agent skill that runs multi-phase security audits with independently verified, machine-checkable findings.security · JEV 0.63OpenAI disrupts Moonshot-linked 'adversarial distillation' campaignOpenAI disclosed it shut down a coordinated campaign by 15,000+ users to extract hidden model reasoning, attributing a…security · JEV 0.58Sandlock 0.8.8: deferred commit for agent sandboxesThe process-based Linux AI sandbox (no container, no VM) ships deferred commit: every run returns a changeset of what…security · JEV 0.56ClawSecure: free security scanner for OpenClaw AI agent skillsFree scanner that audits OpenClaw agent skills for vulnerabilities; the maker's audit of 2,890+ OpenClaw skills found…security · JEV 0.5Google DeepMind SynthID BioWatermarking technology that embeds an imperceptible, verifiable signature into AI-designed protein sequences and…security · JEV 0.49GitHub Security Lab Taskflow Agent: autonomous LLM fuzzing for C/C++Open-source agent that automates the full fuzzing lifecycle - entry-point discovery, harness generation, AFL++ runs…security · JEV 0.47
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.