English · 7 min read
Claude Opus 5.5 vs GPT-6 Sol vs GPT-6 Luna: which one should a small business use?
Anthropic and OpenAI both released new models on 22 September 2026. Claude Opus 5.5 costs $4 / $20 per 1M tokens, GPT-6 Sol $2 / $10 and GPT-6 Luna $0.10 / $0.50. Here is how they compare on price, context and sourced benchmarks, and which to use for which job.
By Dapols ·
The short answer: use GPT-6 Luna for cheap, high-volume work such as sorting emails, tagging tickets and drafting short replies. Use Claude Opus 5.5 for coding and long multi-step tasks, where it is the strongest of the three and an independent index ranks it first overall. GPT-6 Sol sits in between: cheaper than Opus 5.5 and stronger than Luna, a good default if you already build on OpenAI. With both OpenAI models, keep prompts under 272K tokens, because above that the whole request costs more.
Every number below comes from the lab's own page or from Artificial Analysis, an independent benchmark site, and is labelled with who measured it. Checked on 23 September 2026. Prices are list prices per 1 million tokens on each lab's own API.
Update, 1 October 2026: OpenAI released GPT-6.1 Sol on 29 September at the same $2 / $10 as GPT-6 Sol, with cached input halved to $0.10. Artificial Analysis (independent) scores it 52 against GPT-6 Sol's 48, so wherever this page recommends GPT-6 Sol, GPT-6.1 Sol is now the better pick at the same price. OpenAI has also published comparisons of GPT-6.1 Sol against Opus 5.5 (OpenAI's own figures). Our best AI model for coding guide covers it.
The three models side by side
| Claude Opus 5.5 | GPT-6 Sol | GPT-6 Luna | |
|---|---|---|---|
| Lab | Anthropic | OpenAI | OpenAI |
| Released | 22 September 2026 | 22 September 2026 | 22 September 2026 |
| API model id | claude-opus-5-5 | gpt-6-sol | gpt-6-luna |
| Price, input / output | $4 / $20 | $2 / $10 | $0.10 / $0.50 |
| Cached input | $0.20 | $0.20 | $0.01 |
| Prompts over 272K tokens | Same price across the full 1M window | $4 / $15 for the whole request | $0.20 / $0.75 for the whole request |
| Context window | 1M tokens | 1,050,000 tokens | 1,050,000 tokens |
| Max output | 128K tokens | 128,000 tokens | 128,000 tokens |
| Artificial Analysis Intelligence Index (independent) | 58, ranked 1st of 212 | 48, ranked 18th of 212 | 37, ranked 6th of 183 in its class |
| Source | anthropic.com, models overview | developers.openai.com | developers.openai.com |
The index scores are from Artificial Analysis on 23 September 2026, at each model's highest effort setting. On the same scale, GPT-6 Astra and Claude Fable 5.1, the two $10 / $50 flagships, both score 53.
What each lab reports
These are the labs' own figures, not independent tests.
- Claude Opus 5.5 (Anthropic): 66.4% on Terminal-Bench 4.0, against 55.8% for Claude Fable 5.1 and 52.3% for Opus 5. 57.8% on CursorBench 4.0. 1846 Elo on GDPval-AA v2.1, a test of everyday knowledge work. Source.
- GPT-6 Sol (OpenAI): 33.2% on AutomationBench, a test of business workflows across apps, at $0.27 per task. OpenAI says Claude Opus 5 scored 26.9% at 11.1 times the cost. 68.8% on DeepSWE v1.1 at max effort. Source.
- GPT-6 Luna (OpenAI): 66.6% on DeepSWE v1.1 at max effort, which OpenAI calls comparable to Claude Opus 5 and Fable 5 at medium effort, for 93% and 96% less per task. Source.
OpenAI's comparisons are against Claude Opus 5, not Opus 5.5, because Opus 5.5 came out the same day. No lab has yet published a head-to-head of Opus 5.5 against Sol or Luna.
Which one for which job
| Job | Our pick | Why |
|---|---|---|
| Sorting, tagging, short replies, summaries at volume | GPT-6 Luna | One of the lowest list prices from a major lab, below DeepSeek V4.1 Flash off-peak |
| The same volume work, but with very long documents | DeepSeek V4.1 Flash, off-peak | Over 272K tokens Luna costs $0.20 / $0.75; DeepSeek V4.1 Flash off-peak is $0.15 / $0.60 |
| Coding, and agents that run many steps on their own | Claude Opus 5.5 | Highest reported coding scores of the three, and first on the independent index |
| General assistant work on an OpenAI stack | GPT-6 Sol | Half the price of GPT-5.6 Sol, and the same API you already call |
| Customer data that must not leave your own servers | None of these three | All three are cloud-only; use an open-weights model such as DeepSeek V4.1 Flash on your own hardware |
Example, not a quote: 1,000 customer replies, each with 2,000 tokens of input and 500 tokens of output, at list price and without caching, cost about $0.45 on Luna, $9 on Sol and $18 on Opus 5.5. Reasoning tokens are billed as output, so a model that thinks longer costs more than this. Token counts also differ between labs: Anthropic says its tokenizer since Claude Opus 4.7 produces about 30% more tokens for the same text, so compare real bills from a test run rather than rate cards alone.
Things to check before you switch
- The 272K rule on OpenAI. A prompt over 272K input tokens is billed at twice the input rate and 1.5 times the output rate, for the whole request, not just the extra part. If your workflow pastes whole contracts or long chat histories, measure prompt size first.
- Old model ids cost double.
gpt-5.6-solandgpt-5.6-lunastill work, at twice the price ofgpt-6-solandgpt-6-luna. GPT-5.6 Sol is on a promotional $4 / $20 that OpenAI says runs at least until 21 November 2026. - Older Claude models are legacy, not gone. Anthropic now lists Opus 5, Fable 5 and Opus 4.8 as legacy models. They still work. Moving to Opus 5.5 is a one-line model id change, and it is cheaper than Opus 5.
- In ChatGPT, not all plans get the same thing. GPT-6 Sol and Luna are in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu. Free and Go users get Luna in the desktop app. OpenAI says they are not yet in regular ChatGPT chat.
Our model tracker keeps the current pick per category, with sources. If you want a stack chosen for your own workflow rather than a general ranking, that is what a Business AI Plan does. The free 2-minute AI plan finder is the quick first step.
Frequently asked questions
Is Claude Opus 5.5 better than GPT-6 Sol? On the independent Artificial Analysis Intelligence Index, yes: Opus 5.5 scores 58 and GPT-6 Sol 48. Opus 5.5 also costs twice as much per token, at $4 / $20 against $2 / $10.
How much does GPT-6 Luna cost? $0.10 per 1M input tokens and $0.50 per 1M output tokens, with cached input at $0.01. A prompt over 272K input tokens is billed at $0.20 / $0.75 for the whole request.
When were Claude Opus 5.5, GPT-6 Sol and GPT-6 Luna released? All three on 22 September 2026.
Is GPT-6 Sol the same as GPT-5.6 Sol? No. GPT-6 Sol is the newer model and costs half as much. GPT-5.6 Sol, released on 9 July 2026, is still sold at a promotional $4 / $20.
Which is cheapest for a small business? GPT-6 Luna, for prompts under 272K tokens. For longer prompts, DeepSeek V4.1 Flash at off-peak hours is cheaper.
Can I run any of these on my own server? No. All three are cloud-only. If data has to stay in-house, use an open-weights model such as DeepSeek V4.1 Flash.
Sources: Anthropic — Claude Opus 5.5, Anthropic — models overview, Anthropic API pricing, OpenAI — Introducing GPT-6 Sol and Luna, OpenAI — GPT-6 Sol model page, OpenAI — GPT-6 Luna model page, OpenAI API pricing, Artificial Analysis — Claude Opus 5.5, Artificial Analysis — GPT-6 Sol, Artificial Analysis — GPT-6 Luna, DeepSeek pricing. Checked 23 September 2026.