summree
Everything you NEED to know about GPT-5.6
← OpenAI
Matthew Berman

Everything you NEED to know about GPT-5.6

⏱ 9 min video · 2 min read9 Jul 2026Worth watching
TL;DR
Matthew Berman reviews GPT-5.6, OpenAI's latest model family (Luna, Terra, Sol), demonstrating its exceptional browser and computer-use capabilities through week-long autonomous coding runs that produced a functional Excel clone and a Minecraft clone. He compares it to the newer 'Fable' model and covers pricing, benchmarks, and multi-model routing strategies using Codex.
Key points
1
GPT-5.6 comes in three sizes — Luna (small), Terra (medium), Sol (large) — each with multiple reasoning levels up to Ultra, enabling granular cost/performance tradeoffs.
2
A single /goal prompt ran Codex autonomously for 5+ days to build a functional Excel clone with pivot tables, conditional formatting, and data validation using computer use to reference actual Excel.
3
A Minecraft clone was built in ~7 days with realistic 3D rendering, biomes, mobs, and inventory — described as the best AI-generated Minecraft the creator has ever seen.
4
GPT-5.6 is significantly cheaper than GPT-5.5 Fable: $5 vs $10 per million input tokens, $30 vs $50 per million output tokens, and uses fewer tokens to achieve the same result.
5
Box AI benchmarks show Sol outperforming GPT-5.5 (63.3%) and Terra in accuracy on enterprise knowledge work, with Luna matching Terra at lower cost and speed.
Actionable insights
→
Use Sol for planning, Terra on high-reasoning implementation tasks, and Luna for low-requirement work like deployment to stretch Codex quota without sacrificing quality.
→
The creator published a GitHub skill for automated model routing within Codex — download it to delegate tasks across Luna/Terra/Sol automatically.
→
For browser and computer-use tasks (email triage, DNS changes, web automation), GPT-5.6 via Codex browser is described as a production-ready default driver worth adopting now.
Notable quotes

“5.6 feels like the absolute pinnacle of an existing model, where Fable feels like we're just scratching the surface on what's possible on a brand new training run.”

“GPT 5.6 is the most souped-up Honda Civic you've ever seen. Every single horsepower has been squeezed out of it. And Fable is like a Ferrari that hasn't been touched yet, fresh off the manufacturing line, unoptimized, and the potential is just so much higher there.”

“It used computer use. It opened up Excel on my desktop and would just go back and forth between doing something in Excel, the actual Excel, and then recreating it in this new cloned version.”

Worth watching?
✅
Worth watching the full video?
Watch if you want to see the Excel and Minecraft demo footage in action — the key specs and strategy are all captured here, but the live autonomous-agent builds are genuinely impressive to see running.
Topics
AI & TechOpenAI

Explore more summaries on these topics →

Saved you some time? The creator still deserves a like.

Watch on YouTube →
More like this
"The best thing since OpenClaw" (Hermes Tutorial)
Matthew Berman
"The best thing since OpenClaw" (Hermes Tutorial)
Matthew Berman walks through a full setup and feature tour of Hermes Agent, an AI agent platform hosted on Hostinger, positioning it as a more capable and user-friendly alternative to OpenClaw. The video covers installation in under 2 minutes, skills, scheduled tasks, memory, multi-provider model routing, plugins, and Telegram integration.
2 min · 30 Jun 2026
Cut your AI cost IN HALF (EASY)
Matthew Berman
Cut your AI cost IN HALF (EASY)
Matthew Berman explains model routing — using expensive frontier AI models only for planning and spec-writing, then delegating actual code execution to cheaper models. This simple workflow change can cut AI costs by 60-70% with minimal quality loss.
3 min · 7 Jul 2026
We tested OpenAI's GPT 5.6 Sol for 30 days
Greg Isenberg
We tested OpenAI's GPT 5.6 Sol for 30 days
Dan Shipper demonstrates how he uses OpenAI Codex with GPT-5.6 as a full operating system for his personal and business life, showing live builds of an email triage app, a company pulse feed, and a SaaS maintenance badge tool called Turnaround. The episode covers practical Codex patterns, the business opportunity in agent-native SaaS, and the pirate vs architect builder framework.
3 min · 9 Jul 2026

Found this by searching? That's the hard way.

summree watches the AI channels you follow and sends you the summary the moment a video drops. Nothing gets past you. From £4/month.

Try it free