Breaking

OpenAI Adds GPT-6 Sol and GPT-6 Luna as Lower-Cost Agent Models Enter Arena Testing

September 22, 2026 at 14:48 EDT

Foundation Models
AI Agents
Software Dev & Coding

OpenAI has added GPT-6 Sol and GPT-6 Luna to its GPT-6 lineup, positioning the pair as lower-cost models for coding and agentic work, while Arena has begun collecting Agent Arena votes before publishing scores. The launch turns GPT-6 from an Astra-led flagship release into a broader family aimed at price-sensitive, long-running workflows.

Sep 22, 2026 | OpenAI

GPT-6 gets its cost-efficient tiers

GPT-6 Sol and GPT-6 Luna bring lower API prices, long-context agent tooling and new Agent Arena testing before public scores arrive.

50%

API price reduction versus GPT-5.6 promotional pricing

$2 / $10

GPT-6 Sol per 1M input and output tokens

$0.10 / $0.50

GPT-6 Luna per 1M input and output tokens

1.05M

Token context window for both models

Output price falls by half or more

Stack height is proportional to output-token price within each model tier.

Sol output price

$20

GPT-5.6 Sol

$10

GPT-6 Sol

Luna output price

$1.20

GPT-5.6 Luna

$0.50

GPT-6 Luna

How the story moves from launch to evidence

API launch

Tool-use rollout

Agent Arena votes

Scores pending

Why developers care

Lower per-task pricing gives teams more room to run longer coding agents, retry workflows and reserve Astra for the hardest jobs.

What remains unresolved

Agent Arena scores are not published yet, rollout is gradual and reasoning levels can make tier selection more complex.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.