Leaked Gemini 3.8 Flash benchmarks reportedly beat Opus 5 on Terminal Bench, HLE
kimmonismus · x · 2026-09-02
- An unverified leak claims Gemini 3.8 Flash outperforms Claude Opus 5 and others on Terminal Bench 2.1, HLE, and more.
- Follow-up discussion flags its pricing as expensive for the tier.
- Details remain rumor until Google confirms.
Related event: Gemini 3.8 Flash Appears to Be Shadow-Launching Ahead of Imminent Release(33 posts)→
More from Models
- Google ships Gemini 3.8 Flash, third Flash update in 6 weeks, boosting agentic and coding skills — reach_vb · 2026-09-03
- Google launches Gemini 3.8 Flash Cyber security model alongside Fairwind Program for defenders — 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
- Early Gemini 3.8 testing suggests Gemma 4.5 will be very good — andrew_n_carr · 2026-09-03
- OpenAI's Astra model checkpoints spotted: vega-alpha and ultima-alpha in testing — koltregaskes · 2026-09-03
- Code Arena launches WebDev Pareto frontier to rank AI models by quality vs price — arena · 2026-09-03