GPT-6 Sol and Luna Are HERE!

Summary
Matthew Berman reviews OpenAI’s release of GPT-6 Sol and GPT-6 Luna, comparing them to recent model releases including Anthropic’s Claude Opus 5.5 and xAI’s Grok 4.7. He walks through official benchmark charts and API pricing data to evaluate their cost efficiency across agentic coding, computer use, and general automation tasks.

What is shown

• [00:00] Introduction of GPT-6 Sol and GPT-6 Luna against backdrop graphics.
• [00:22] OpenAI blog announcement and social media posts from OpenAI’s Thibault Sottiaux (@thsottiaux).
• [00:41] GPT-6 API pricing table displaying token rates and percentage discounts compared to GPT-5.6 Sol and Luna.
• [01:30] Sponsor presentation for here.now website and widget deployment for AI coding agents.
• [02:21] AutomationBench chart analyzing task scores against cost per task for GPT-6 Astra, Sol, Luna, and Claude models.
• [03:49] Agents’ Last Exam benchmark scatter plot comparing cost and score.
• [03:59] FrontierCode V1.1 Main benchmark scatter plot comparing coding capabilities across cost levels.
• [04:18] DeepSWE agentic software engineering benchmark graph.
• [04:36] OSWorld 2.0 offline set benchmark evaluating computer-use performance.
• [04:54] Side-by-side comparison bar charts comparing pricing, FrontierCode, and AutomationBench across Luna, Sol, and Claude Opus 5.5.
• [05:32] OpenAI Availability announcement details for ChatGPT Work, Codex, and API access tiers.

Claims & numbers

• The presenter notes that GPT-6 Sol and Luna launched on the same day as Anthropic’s Claude Opus 5.5 and one day after xAI’s Grok 4.7.
• API pricing for GPT-6 Sol is $2 per million input tokens and $10 per million output tokens, reflecting a 50% cut from GPT-5.6 Sol ($4 / $20).
• API pricing for GPT-6 Luna is $0.10 per million input tokens and $0.50 per million output tokens, a 50% cut from GPT-5.6 Luna ($0.20 / $1.20).
• On AutomationBench, the presenter states GPT-5.6 Sol scored 28.8% at $0.67 per task, GPT-6 Sol scored 33.2% at $0.27 per task, GPT-6 Astra scored 41.4% at $1.73 per task, and GPT-6 Luna scored 20.7% at $0.037 per task (max effort) or 1.2% at $0.006 per task (low effort).
• On Agents’ Last Exam, GPT-6 Astra reached 59.3% ($6.23/task), while GPT-6 Sol reached 56.4% ($2.93/task at max effort) and 55.4% ($1.67/task at extra-high effort).
• On FrontierCode V1.1 Main, GPT-6 Luna reached 42.4% at $0.11 per task, GPT-6 Sol scored 45.9% at $0.80 per task, and GPT-6 Astra scored 45.3% at $1.70 per task.
• On DeepSWE, the presenter highlights GPT-6 Luna Max reaching a 66.6% score at $0.22 cost per task.
• On OSWorld 2.0 offline set, GPT-6 Luna Max reached 52.7% ($0.27/task) while GPT-6 Astra reached 73.5% ($9.07/task).
• The presenter states GPT-6 Sol and Luna are rolling out to ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, with desktop app access for Free/Go users and API availability as gpt-6-sol and gpt-6-luna.

Notable quotes

• [00:46] “GPT-5.6 Sol versus GPT-6 Sol: a 50% price reduction.”
• [01:21] “Luna can probably take on 90 to 95% of all the tasks that you have to give it.”
• [06:23] “You get efficiency, you get speed, you get price, but you don’t get absolute raw power and intelligence.”

Assessment
This is an independent analysis and review summarizing OpenAI’s official release notes, pricing, and benchmark figures. The presenter does not run interactive live tests on camera, relying instead on official benchmark slides, charts, and public social media statements.

Described by gemini-3.8-flash on 2026-10-12 from the video’s audio and frames.