Claude vs ChatGPT 2026: Fable 5.1 and Opus 5.5 vs GPT-6

If you are comparing Claude and ChatGPT using August figures, most of those figures are now wrong. Both vendors changed their lineups in September.

Anthropic released three models this month: Claude Fable 5.1 on 1 September, Claude Opus 5.5 on 22 September and Claude Sonnet 5.5 on 28 September. OpenAI announced GPT-6 Sol and GPT-6 Luna on 22 September, and its model documentation now lists GPT-6 Astra, Sol and Luna first. GPT-5.6 Sol and Claude Opus 5, which most comparisons still use, have both been replaced.

This is the updated comparison. We cover the paid API and the free chat apps separately because the answers differ.

Which Is Better in 2026, Claude or ChatGPT?

For most paid work, it depends on which tier you use. For hours-long agentic work, choose Claude. For the lowest price at volume, choose ChatGPT.

The flagships cost the same. Claude Fable 5.1 and GPT-6 Astra are both $10 per million input tokens and $50 per million output, so price no longer decides between them. Anthropic says to use Fable 5.1 for demanding reasoning and long-horizon agentic work. It also says to start with Claude Opus 5.5 for most workloads and to move up to Fable 5.1 only if your evaluations on Opus 5.5 at higher effort still fall short. That advice tells you what each model is for.

Lower down the lineup, OpenAI is cheaper. GPT-6 Sol costs $2 and $10, half the price of Opus 5.5. GPT-6 Luna costs $0.10 and $0.50. If most of your work is summarising, extraction and routine drafting, those prices will matter more than any difference in quality.

What Changed for Both This Month?

On Anthropic's side, Claude Opus 5.5 was released on 22 September 2026 and is now the model Anthropic's models overview tells you to start with. It has a 1M token context window, 128K max output, and costs $4 input and $20 output, which is cheaper than Opus 5 at $5 and $25. Claude Fable 5.1, released 1 September, is still the top of the lineup at 1M context and $10 and $50. Claude Sonnet 5.5 arrived on 28 September at $2 and $10, and Haiku 4.5 is still the fast, low-cost model.

If you are upgrading existing code, expect to change more than the model name. Anthropic lists four breaking changes from Opus 5 to Opus 5.5. Thinking can no longer be disabled. Forced tool use returns an error. Thinking blocks are tied to the model and conversation that produced them. On the Claude API and Google Cloud, the older computer use tool version is no longer accepted.

On OpenAI's side, the OpenAI model documentation now lists GPT-6 Astra as the most capable model, GPT-6 Sol for complex coding and agentic workflows, and GPT-6 Luna for focused, high-volume tasks. According to OpenAI's 22 September announcement, Sol and Luna launched in the API and in ChatGPT Work and Codex for paid plans. Free and Go users can try GPT-6 Luna in the desktop app.

Our Claude review and ChatGPT review cover each product in more depth.

How Do the Context Windows Compare Now?

On paper they are close to equal, and the small difference rarely matters.

OpenAI documents all three GPT-6 models at 1,050,000 tokens of context, with a 922,000 token limit on input and 128,000 on output. Claude Fable 5.1, Opus 5.5 and Sonnet 5.5 are 1M tokens, with 128K max output. Haiku 4.5 is 200K. Anthropic's 1M covers the whole window with no separate input limit, so the space available for input is closer than the headline numbers suggest.

The Claude chat apps now offer the same large window. Anthropic's pricing page lists the context window as up to 1M, varying by model, on Free, Pro and Max. Its help centre article on paid plans gives 1M in chat for Fable 5.1, Opus 5.5, Opus 5, Sonnet 5.5 and Sonnet 5, and 500K for several older models. That article covers paid plans only, so we can't give a free user an exact figure, but the old claim that Claude chat stops at 200K on every plan is no longer true.

What Do They Actually Cost to Run?

At the top they cost the same. Below that, OpenAI is cheaper.

Per million tokens, input and output:

• Top tier: Claude Fable 5.1 at $10 and $50, GPT-6 Astra at $10 and $50.
• Everyday tier: Claude Opus 5.5 at $4 and $20, GPT-6 Sol at $2 and $10. Claude Sonnet 5.5 is also $2 and $10.
• Budget tier: Claude Haiku 4.5 at $1 and $5, GPT-6 Luna at $0.10 and $0.50.

Luna costs a tenth of Haiku 4.5. It is not a like-for-like model comparison, but for high-volume classification and extraction, a tenfold price difference will usually outweigh any quality gap.

Caching changes the numbers at the top. Fable 5.1 cache reads cost $0.25 per million, while GPT-6 Astra's cached input costs $1. A workload with a large, stable prompt prefix that is reused heavily will cost noticeably less on Fable 5.1. In the middle the cache prices are the same: Opus 5.5 and GPT-6 Sol both charge $0.20. OpenAI also charges 2x input and 1.5x output on the whole request once input goes above 272K tokens, so very large prompts cost more than the list price.

Which One Wins on the Free Tier?

ChatGPT wins on volume. Claude has fewer free models, but its limit resets every five hours.

OpenAI gives free users unlimited everyday text chats on GPT-5.6 Luna, with limits on file uploads, images and other tools. As of 22 September, free users can also try GPT-6 Luna in the desktop app. If you mostly type questions and read answers, you are unlikely to hit a limit.

Claude Free has fewer models than we reported before. The pricing page lists Sonnet and Haiku on Free, but not Opus or Fable. Opus starts on Pro, and Fable needs usage credits on Pro or comes with Max. Usage on Free resets on a rolling five-hour window, and Anthropic gives no fixed message count. The weekly limits on the pricing page apply to paid plans. Pro gives at least 5x more usage per five-hour session than Free. Free also includes up to 5 Projects, and the context window is listed as up to 1M, varying by model.

If you want the most free usage, choose ChatGPT. If you want Claude's writing and document reading and can work within a five-hour reset, Claude Free is a good option. Our sister site keeps a running list of genuinely free AI tools if neither free plan suits you.

Which Is Better for Writing and Long Documents?

Claude. The new models have not changed our view here.

Claude still writes more natural prose, keeps a consistent voice across a long piece, and follows multi-part instructions more precisely. With long documents it reads more carefully, and context window numbers don't capture that. A model that accepts a million tokens but skims them is worse for contract review than one that reads every clause.

Context size no longer separates them. Both vendors document roughly a million tokens on their current models, so the old reason to pick ChatGPT for very large documents has mostly gone. The main reason left to choose ChatGPT for document work is cost: running a large archive through GPT-6 Luna at $0.10 per million input tokens is much cheaper than any Claude model.

For better results from either model on writing tasks, how you phrase the prompt matters more than which model you use. Structured prompt patterns are collected at PromptCraft Asia, and our Claude vs ChatGPT writing comparison tests both on specific writing tasks.

Do Both Work Properly From Southeast Asia?

Yes. Both work without a VPN in Singapore, Malaysia, the Philippines, Indonesia and Thailand.

Neither vendor restricts access by region, and both accept Singapore-issued cards on paid plans. Claude Pro costs $20 a month billed monthly or $17 a month billed annually, roughly SGD 27 and SGD 23. Max starts at $100 a month. ChatGPT Plus is $20 a month, and ChatGPT Go at $8 is a cheaper step up from free.

Both store conversations in the United States. For most work that doesn't matter. For regulated material, check data residency before you look at any of the numbers above. It is also the reason both still beat cheaper China-hosted options for business use.

Which Should You Pick for Your Own Work?

These four rules cover almost everyone.

Choose GPT-6 Sol if you are building on the API and want one low-cost default for general professional work. It costs half as much as Opus 5.5, and it is the same price as Claude Sonnet 5.5, so compare those two on your own tasks.

Choose Claude Opus 5.5, or Fable 5.1 if your evaluations need it, if your agent runs unattended for hours, does multistep research, or produces documents and spreadsheets. At the top tier, Fable 5.1 and GPT-6 Astra cost the same, so decide on quality rather than price.

Choose ChatGPT Free if you don't want to pay and you type a lot. Unlimited everyday text chats is the most generous free-tier feature available.

Choose Claude Free if you don't want to pay and answer quality matters more to you than volume. You get Sonnet and Haiku with a limit that resets every five hours. You'll need Pro for Opus, and Pro with usage credits or Max for Fable.

If you can't decide, try one real task in both this week. A couple of hours of your own testing will tell you more than any comparison table, including this one.

What Else Do People Ask?

Which Claude and ChatGPT models are current in September 2026?

On Anthropic's side: Claude Fable 5.1 (released 1 September 2026), Claude Opus 5.5 (22 September), Claude Sonnet 5.5 (28 September) and Claude Haiku 4.5. Anthropic tells you to start with Opus 5.5 for most workloads and to use Fable 5.1 for demanding reasoning and long-horizon agentic work. On OpenAI's side, the model documentation now leads with GPT-6 Astra, GPT-6 Sol and GPT-6 Luna. Sol and Luna were announced on 22 September. The GPT-5.6 family and Claude Opus 5 are no longer the current lineups.

Does ChatGPT have a bigger context window than Claude now?

On the API, marginally. OpenAI documents GPT-6 Astra, Sol and Luna at 1,050,000 tokens of context with 922,000 maximum input tokens and 128,000 output tokens. Claude Fable 5.1, Opus 5.5 and Sonnet 5.5 are 1M tokens, and Haiku 4.5 is 200K. In the Claude apps, Anthropic's pricing page lists the context window as up to 1M, varying by model, on every plan including Free, and its help centre documents 1M in chat on paid plans for Fable 5.1, Opus 5.5, Opus 5, Sonnet 5.5 and Sonnet 5.

How much do the Claude and GPT-6 models cost per million tokens?

At the top the price is identical: Claude Fable 5.1 and GPT-6 Astra both cost $10 per million input tokens and $50 per million output. Below that OpenAI is cheaper. Claude Opus 5.5 is $4 and $20, against GPT-6 Sol at $2 and $10, the same price as Claude Sonnet 5.5. GPT-6 Luna is $0.10 and $0.50, a tenth of Claude Haiku 4.5 at $1 and $5. OpenAI charges 2x input and 1.5x output on the whole request above 272K input tokens.

Which has the better free tier, Claude or ChatGPT?

ChatGPT, on volume. OpenAI gives free users unlimited everyday text chats on GPT-5.6 Luna, with limits on file uploads, images and other tools, and free users can try GPT-6 Luna in the desktop app. Claude Free gives you Sonnet and Haiku, not Fable or Opus, with usage that resets on a rolling five-hour window rather than a weekly cap. Pro gives at least 5x more usage per five-hour session. Free also includes up to 5 Projects, and the context window is listed as up to 1M, varying by model.

Which should you choose for long document work?

For most people, Claude. Both vendors now document roughly a million tokens of context on their current API models, so raw capacity no longer separates them, and Claude still reads long documents more carefully and holds a consistent voice across long output. Choose ChatGPT if you need the cheapest possible pass over very large volumes, where GPT-6 Luna at $0.10 per million input tokens is hard to argue with.

Sources: Anthropic's models overview for the current lineup, the context windows and API prices, and the advice to start with Opus 5.5. The model pages for Claude Opus 5.5 (22 September 2026 release, $0.20 cache reads, breaking changes), Claude Sonnet 5.5 (28 September 2026 release) and Claude Fable 5.1 (1 September 2026 release, $0.25 cache reads). Anthropic's pricing page for which models each plan includes, the rolling five-hour limit window, up to 5 Projects on Free, the up to 1M context window on every plan, and Pro at $20 monthly or $17 annual. Anthropic's help centre for context windows by model in chat on paid plans. OpenAI's model documentation and its GPT-6 Astra, GPT-6 Sol and GPT-6 Luna pages for context, input and output limits, per-token and cached pricing, and the 272K long-context surcharge. OpenAI's 22 September 2026 announcement for GPT-6 Sol and Luna availability, including GPT-6 Luna for free users in the desktop app. OpenAI's ChatGPT Free Tier FAQ for unlimited everyday text chats on GPT-5.6 Luna and the remaining limits on uploads, images and tools. All figures were checked on 29 September 2026 and may change, so check with each vendor before you choose a plan.