OpenAI’s new flagship, GPT-5.6 Sol, is a reasoning model built for agentic coding, and it just set a new SOTA on Terminal-Bench 2.1 — the benchmark that measures planning, iteration, and tool-juggling inside a real command line. Base Sol scores 88.8%. The “Sol Ultra” config pushes that to 91.9%, ahead of Claude Mythos 5 (84.3%) and Gemini 3.1 Pro Preview (70.7%).
Two things are actually new. A max reasoning effort mode lets the model think longer on hard problems. And ultra mode dispatches subagents to run work in parallel, breaking past what a single agent can do. That second one is why developers care: a tweet reading “GPT-5.6 Sol Ultra will be in Codex” hit 287 points on HackerNews today.
What you can actually do with it
Sol runs through the API and inside Codex, so ultra’s multi-agent execution becomes your coding assistant. Pricing: $5 input / $30 output per million tokens.
The catch
It’s a locked preview — roughly 20 government-vetted orgs only. OpenAI cleared release plans with the U.S. government first. Everyone else waits.
You Might Also Like
- Hiveterm Bets on the Multi Agent Workspace Claude Codex and Gemini in one Terminal
- Openai Codex Pets Turn Your ai Coding Agent Into a Desktop Tamagotchi
- Openai Codex in Chrome Moves the Coding Agent Into Your Real Browser Session
- Openai gpt 5 6 sol Terra Luna a Three Tier Lineup Only 20 Orgs can Touch
- Openai Codex Record and Replay Demo a mac Task Once the Agent Repeats it Forever
Top AI Product 
Leave a comment