Google Ships Gemini 3.8 Flash: 1M Context, Claims Terminal Bench Wins Over Rival Flagships
kimmonismus · x · 2026-09-02
Google has released Gemini 3.8 Flash, its most capable Flash model aimed at long-horizon software engineering, autonomous agents, and enterprise workflows.
- Context: 1,048,576-token input, 65,536-token output
- Capabilities: text/image/video/audio/PDF input; Thinking (low/medium/high, minimal unsupported); code execution; Computer use (Preview); Search and Google Maps grounding; structured outputs, caching, file search; Batch/Flex/Priority inference; no audio/image generation or Live API
- Benchmarks: poster kimmonismus claims Flash beats competing flagships on Terminal Bench 2.1, HLE and more (third-party reading, not item-by-item verified by Google)
Related event: Google Launches Gemini 3.8 Flash and Cybersecurity-Focused Flash Cyber(57 posts)→
More from Models
- Schmidhuber Says OpenAI's 'Recurrent Depth' Reasoning Echoes His 2015 Learning-to-Think Paper — SchmidhuberAI · 2026-09-03
- Google launches Gemini 3.8 Flash Cyber security model alongside Fairwind Program for defenders — GoogleAI · 2026-09-03
- Google launches Gemini 3.8 Flash Cyber, helping Chrome team produce 2.6x more correct patches — GoogleAI · 2026-09-03
- Matt Shumer: slow AI releases aren't a wall — safety clearance is the bottleneck, wave of frontier models imminent — mattshumer_ · 2026-09-03
- Unverified claim: Gemini 3.8 Flash tops DeepSWE v1.1 coding benchmark — vedantmisra · 2026-09-03
- Early Gemini 3.8 testing suggests Gemma 4.5 will be very good — andrew_n_carr · 2026-09-03