OpenAI Announces GPT-5.6 Series, Flagship Model Surpasses Claude Mythos 5
The GPT-5.6 series marks a clear performance leap over the previous generation and over Anthropic's current flagship, but the U.S. government's intervention in limiting initial access raises questions about how future frontier models will be distributed.
Reporting from 1 sources: GIGAZINE.
OpenAI announced the GPT-5.6 series on June 26, 2026, comprising three models: flagship GPT-5.6 Sol, balanced GPT-5.6 Terra, and low-cost GPT-5.6 Luna. GPT-5.6 Sol outperformed Anthropic's Claude Mythos 5 on Terminal-Bench 2.1 and matched it on ExploitBench at one-third the output tokens. The entire series is currently limited to a small group of trusted partners at the U.S. government's request, with public release expected within weeks.
OpenAI unveiled the GPT-5.6 series on June 26, 2026, a three-model lineup led by GPT-5.6 Sol, which beat Anthropic's Claude Mythos 5 on the Terminal-Bench 2.1 benchmark for practical terminal tasks. On ExploitBench, which measures cyberattack capability, Sol matched Mythos Preview while using a third of the output tokens. The mid-range Terra scored comparably to Claude Fable 5, and the budget Luna surpassed Claude Opus 4.8 and Gemini 3.1 Pro Preview.
Pricing per million tokens ranges from $5 input and $30 output for Sol down to $1 input and $6 output for Luna. OpenAI said it briefed the U.S. government before the announcement and limited the preview to a small set of trusted partners at the government's request. The company called the arrangement a short-term measure and said it plans a broader public release within weeks while working with the administration on a framework for future model releases.
- GPT-5.6 Sol: flagship model, highest performance, $5/$30 per million tokens
- GPT-5.6 Terra: balanced model for everyday tasks, $2.50/$15 per million tokens
- GPT-5.6 Luna: low-cost model, $1/$6 per million tokens
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.