OpenAI Unveils GPT-5.6 Series, Outperforming Claude Mythos 5
OpenAI officially launched its GPT-5.6 model series on June 27, featuring three tiers: the flagship Sol, balanced Terra, and lightweight Luna. The flagship Sol scores 88.8% on the Terminal-Bench 2.1 coding benchmark, surpassing Anthropic's Claude Mythos 5, and reaches 91.9% with Ultra mode enabled, with notably improved token efficiency in cybersecurity reasoning tasks. Under US government regulatory requirements, the series is currently only available to approximately 20 "trusted partners" for preview, with gradual expansion expected in the coming weeks. Priced at $5 per million input tokens and $30 per million output tokens, Sol costs roughly half of Anthropic's high-end models, officially kicking off the LLM price competition.
#GPT56 #OpenAI #LLM #AIRegulation #Claude
