GPT-6.1 Sol vs GPT-6 Luna: Model Tiers, Pricing, and Workload Fit

What OpenAI says Luna and GPT-6.1 Sol are for

OpenAI describes GPT-6 Luna as its most efficient model, meant for focused, high-volume tasks. It describes GPT-6.1 Sol as close to GPT-6 Astra's performance at a lower cost, for complex coding, computer use and professional work. GPT-6.1 Sol takes a reasoning effort setting from low to max; it has no "none" setting. Luna's knowledge cutoff, May 18, 2026, is a little later than GPT-6.1 Sol's April 30, 2026.

Those descriptions do not tell you where Luna stops being good enough for your tasks. A sample run does: send a few hundred real requests to Luna, check the output the way production would, and move only the task types that fail to GPT-6.1 Sol. At a twentieth of the price, Luna can absorb a retry or a second check and still cost less.

Per million tokens, Luna costs a twentieth of GPT-6.1 Sol

These are OpenAI's Standard rates. Read the "over 272K" rows for any request whose input passes 272,000 tokens, because the higher rates then apply to the whole request.

On uncached input and on output the gap is exactly 20 times. Cached input narrows it to 10 times, because GPT-6.1 Sol bills cached input at 5% of its input rate and Luna at 10%. Writing to the cache costs 1.25 times the input rate on both. Above 272,000 input tokens, input and cache rates double and output rises by half, so a 300,000-token prompt to GPT-6.1 Sol costs $1.20 in input instead of $0.60.

Batch and Flex halve both; of the two, only GPT-6.1 Sol has Ultrafast

Batch and Flex cost half of Standard. A Batch job runs under its own, higher rate limits and completes within 24 hours. Fast, which OpenAI called Priority processing until July 30, 2026, costs twice Standard: $4 and $20 for GPT-6.1 Sol, $0.20 and $1 for Luna. Ultrafast costs six times Standard ($12 and $60 for GPT-6.1 Sol) and is offered for GPT-6.1 Sol and GPT-6 Astra, not Luna. Fast Luna still costs a tenth of Standard GPT-6.1 Sol.

The illustrative job below is 1,000 requests, each with 3,000 input tokens and 300 output tokens and no caching; the sizes are ours, not OpenAI's. Read the last column against the Luna row: Batch GPT-6.1 Sol still costs ten times Standard Luna.

So Batch is a way to pay less for work that can wait a day, not a substitute for checking whether Luna can do the task.

In ChatGPT, Go has Luna and Plus adds GPT-6.1 Sol in Work and Codex

ChatGPT plans give you models within usage limits, not tokens. Go ($8 a month) includes GPT-6 Luna but neither Sol model. Plus ($20) includes GPT-6 Sol and GPT-6 Luna, and GPT-6.1 Sol and GPT-6.1 Terra in ChatGPT Work and Codex but not in Chat. On Free and Go, Codex runs Luna at Standard speed in the desktop app as it rolls out.

For local Codex work on Plus, OpenAI estimates 15 to 160 GPT-6.1 Sol messages per five-hour window, against 350 to 3,000 on Luna. GPT-6.1 Terra, added on October 26, 2026, sits between them at 35 to 285 and half of GPT-6.1 Sol's API price; GPT-6.1 Sol vs GPT-6.1 Terra compares that middle step. These are ranges, not fixed limits, and weekly limits may also apply. ChatGPT Go vs Plus covers the plan step, and ChatGPT Work and Codex limits by plan the allowances. Codex signed in with an API key can use GPT-6.1 Sol, GPT-6 Sol or Luna at the API rates above; Codex: ChatGPT login vs API key explains that bill.

GPT-6 Sol is the older Sol; GPT-6 Astra sits above both

GPT-6 Sol still has its own model page, which points to GPT-6.1 Sol as the newer Sol model. OpenAI built it for complex coding and agentic workflows. It costs the same $2 and $10 per million tokens, but its cached input costs $0.20 rather than $0.10, and its knowledge cutoff is April 20, 2026. GPT-6.1 Sol costs the same or less for the same tokens, so keep GPT-6 Sol only where an evaluation is pinned to it. GPT-6.1 Sol vs GPT-6 Sol compares the two.

When GPT-6.1 Sol is not enough, OpenAI's most capable model, GPT-6 Astra, costs $10 and $50 per million tokens, five times GPT-6.1 Sol. See GPT-6 Astra vs GPT-6 Sol.