Google launches Gemini 3.8 Flash and Flash Cyber: 54.9% HLE, 2.6x better vuln patches
AlienX__0 · reddit · 2026-09-21
Google announced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, targeting software engineering, agentic workflows, and cybersecurity.
Gemini 3.8 Flash
- Described as Google's best reasoning and coding model yet, built for long-horizon, multi-step engineering tasks.
- Scores 54.9% on HLE-Verified.
- Introductory API pricing: $0.75/M input tokens, $3.75/M output tokens.
Gemini 3.8 Flash Cyber
- Focused on vulnerability discovery and automated patching; over 70% success on an internal benchmark spanning 20 programming languages.
- Available to trusted defenders via the Fairwind Program.
- Chrome's security team reports 2.6x more correct vulnerability patches than the best commercial models.
More from Models
- Mollick: Claude's missing image generator is a real gap for agentic knowledge work — emollick · 2026-09-21
- GPT-6's 3D hype meets reality: a gorgeous handle you should never print — ngxson · 2026-09-21
- Dev tests model routing: multi-model 'team' costs as much as Astra alone and times out — kevinkern · 2026-09-21
- Diffusion makes inference look like training—the shape GPUs were built for — victor_explore · 2026-09-21
- Developer cancels $200/mo Codex subscription, going all-in on open-source models — haydendevs · 2026-09-21
- Pre-registered test of Jev's no-hallucination claim: type safety holds, calibration splits — NervousAd4440 · 2026-09-21