Skip to main content

Claude vs GPT: Opus 5.5 against GPT-6 Astra

If you searched for Claude vs GPT-4, the comparison has moved. OpenAI removes GPT-4 from its API on 23 October 2026, and both companies now sell a current lineup at matching price points. This page compares what each one sells today: Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 against GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna.

Reviewed against 24 sources. Prices, models, policies and terms change, so check the linked pages before you buy, build, publish or trade.

The short answer

At list price the two lineups cost the same at the top, middle and bottom, so price per token does not settle this. What you send and what you need back does: Claude is the better buy for long prompts and long-running coding agents, GPT for images, voice, a cheaper chat plan and open weights.

Choose Claude if

  • Your prompts run past 272,000 tokens. Claude bills its 1M window at the standard rate; OpenAI doubles the input price above that line.
  • Your agent resends a long prefix every turn. Cached input costs $0.25 per million tokens on Fable 5.1 and $1.00 on GPT-6 Astra.
  • You want the model at the top of independent rankings and will pay more per task for it.

Choose GPT if

  • You need generated images or live voice from the same vendor. Claude models return text.
  • You want near-flagship results on a mid-tier budget: GPT-6.1 Sol is $2 input and $10 output per million tokens.
  • You want the cheapest paid chat plan (ChatGPT Go, $8 a month) or open weights you can host.

The facts, side by side

Each row shows the date it was checked and numbered links to the pages it was read from.

marks the side the facts favour on that row. Rows without one are a draw or a matter of fit.

Swipe the table sideways to see GPT.

Claude and GPT compared, fact by fact
ComparedClaudeAnthropicGPTOpenAI
Who makes it18Anthropic. The chat app is Claude; the developer product is the Claude API.OpenAI. The chat app is ChatGPT; the developer product is the OpenAI API.
Current models18Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna
Where the vendor says to start18Opus 5.5 for most workloads; Fable 5.1 when Opus at higher effort still falls shortGPT-6 Astra for complex reasoning and coding; GPT-6.1 Sol to balance intelligence and cost; GPT-6 Luna for high volume
Context window191M tokens on all four models1,050,000 tokens on all three, with a 922,000-token input limit
Longest single reply19128K tokens128,000 tokens
Knowledge cutoff18Advantage: June 2026 on all four30 April 2026 on Astra and GPT-6.1 Sol; 18 May 2026 on Luna
What goes in and comes out18Text and images in, text outAdvantage: Text and images in, text out on the GPT-6 models. Separate models generate images (GPT-Image-2.5) and hold voice conversations (GPT-Live 1).
Clouds you can buy it through11617Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, Claude Platform on AWSOpenAI API, Microsoft Foundry on Azure, Amazon Bedrock (models vary by AWS Region)
Apps and coding agent372018Claude on web, iOS, Android and desktop. Claude Code in the terminal, IDE, desktop app and browser.ChatGPT on web, mobile and desktop (macOS, Windows, Linux). Codex on the web, CLI, IDE extension and iOS.
Open-weight models115None listedAdvantage: gpt-oss-120b and gpt-oss-20b, under the Apache 2.0 licence
Training on your data63141821API and Team: not used by default. Free, Pro and Max list model training as opt-out.API: not used unless you opt in. Business: not used by default. Personal plans: set by the "Improve the model for everyone" switch.
Notice before a model is retired512At least 60 days. Each current model also has an earliest retirement date; the soonest is 1 September 2027.At least 6 months for generally available models, 3 months for specialised variants

Where each one is stronger

Claude Anthropic

  • Long prompts cost the standard rate

    Fable 5.1, Opus 5.5 and Sonnet 5.5 bill a 900,000-token request at the same per-token price as a 9,000-token one. All three GPT-6 models charge 2x input and 1.5x output on the whole request once the prompt passes 272K tokens. 29

  • Cheaper cache reads on the top model

    An agent that resends the same long prefix every turn pays $0.25 per million cached tokens on Fable 5.1 and $1.00 on GPT-6 Astra. In the middle tier the two match at $0.10. 210

  • First on two independent rankings

    Artificial Analysis scored Claude Opus 5.5 at 58 on its Intelligence Index (version 4.3.2, ten evaluations) at maximum effort, ahead of GPT-6 Astra at 53. Arena's agent leaderboard of 8 October 2026 lists Opus 5.5 first and Astra third, though their rank ranges overlap. 222324

GPT OpenAI

  • Images and voice from one vendor

    OpenAI's catalogue includes GPT-Image-2.5 Sunburst and Flare for image generation and GPT-Live 1 for voice conversations, generally available since 10 September 2026 at $0.05 per minute. Anthropic's current models return text. 8111

  • Close to the top for less per task

    On the same Artificial Analysis index, GPT-6.1 Sol scored 52 at maximum effort for $0.72 per task. Claude Opus 5.5 scored 54 at high effort for $1.82, and its 58 at maximum effort cost $5.98. 22

  • Open weights you can host

    gpt-oss-120b and gpt-oss-20b can be downloaded, fine-tuned and deployed commercially under the Apache 2.0 licence. Anthropic lists no open-weight model. 151

Which to pick, by job

  • A coding agent that works through a large repository for hours

    Pick Claude

    Anthropic describes Opus 5.5 as built for long-running agentic coding, it is first on both independent rankings read for this page, and its 1M window is billed at the standard rate. Start at the default effort: in Artificial Analysis testing, cost per task rose from $1.82 at high effort to $5.98 at maximum. 12224

  • Reading contracts, transcripts or codebases longer than 272,000 tokens

    Pick Claude

    Both lineups accept about a million tokens, but OpenAI bills the whole request at 2x input once it passes 272K. A 600,000-token prompt is $1.20 of input on Claude Sonnet 5.5; the same token count is $2.40 on GPT-6.1 Sol. 210

  • A product that needs generated images or a talking voice

    Pick GPT

    OpenAI sells image and live voice models next to its text models, on one account and one bill. With Claude you would add a second vendor for both. 81

  • A feature that runs thousands of times a day on a fixed budget

    Pick GPT

    GPT-6.1 Sol and Claude Sonnet 5.5 share a list price of $2 and $10, but Sol scored higher for less per task on the Artificial Analysis index: 50 at high effort for $0.32, against 47 for $0.88. Run both on your own prompts before you commit. 22

  • One paid chat app for writing, research and everyday questions

    Pick GPT

    ChatGPT Go is $8 a month and Plus is $20. Claude Pro is $20, or $17 billed annually. Choose Claude Pro instead if you want to write with Opus or work in Claude Code; Pro includes both. 318

What it costs

List prices in US dollars, read from each vendor's own pages. API prices are per million tokens. Token counts depend on each model's tokenizer (Anthropic says its current one produces about 30% more tokens for the same text than its older models), so a per-token price is not a per-task price. Shortlist here, then measure on your own prompts.

marks the side the facts favour on that row. Rows without one are a draw or a matter of fit.

Swipe the table sideways to see GPT.

Claude and GPT prices
ComparedClaudeAnthropicGPTOpenAI
Free plan318Claude Free: $0, with the Sonnet and Haiku modelsChatGPT Free: $0
Lowest paid plan318Claude Pro: $20 a month, or $17 a month billed annually ($200 up front)Advantage: ChatGPT Go: $8 a month. ChatGPT Plus: $20 a month.
Heavy-use plan318Claude Max: from $100 a month, with 5x or 20x the usage of ProChatGPT Pro: $100, $200 or $500 a month
Team seat318Team: $25 per seat a month, or $20 billed annually. Premium seat: $125, or $100 billed annually.Business: $25 per user a month, or $20 billed annually, for 2 or more users
Top API model210Fable 5.1: $10 input, $50 outputGPT-6 Astra: $10 input, $50 output
Mid API models210Sonnet 5.5: $2 input, $10 output. Opus 5.5: $4 input, $20 output.GPT-6.1 Sol: $2 input, $10 output
Cheapest API model210Haiku 5.5: $0.10 input, $0.50 output for prompts up to 100,000 tokens; $0.50 and $2.50 above thatGPT-6 Luna: $0.10 input, $0.50 output
Long prompts29Advantage: Standard rate across the full 1M window on Fable 5.1, Opus 5.5 and Sonnet 5.5Above 272K input tokens: 2x the input price and 1.5x the output price for the whole request
Cached input210Fable 5.1: $0.25. Opus 5.5: $0.20. Sonnet 5.5: $0.10. Haiku 5.5: $0.01.GPT-6 Astra: $1.00. GPT-6.1 Sol: $0.10. GPT-6 Luna: $0.01.
Batch jobs21050% off input and output50% off, on Batch and Flex processing

What changed recently

  1. Scheduled: OpenAI removes gpt-4 and gpt-4-turbo from its API. The named replacement is gpt-5.6-sol. 12

  2. Scheduled: GPT-5.5 retires from ChatGPT and Codex on every plan. The API is not affected. 19

  3. Anthropic launched Claude Haiku 5.5 and cut Sonnet 5.5 cache reads from $0.20 to $0.10 per million tokens. 4

  4. OpenAI cut its API usage tiers from five to three: Build, Launch and Grow. 11

  5. OpenAI released GPT-6.1 Sol at $2 input and $10 output per million tokens. 11

  6. Anthropic launched Claude Sonnet 5.5. 4

  7. Anthropic launched Claude Opus 5.5 at $4 and $20, down from $5 and $25 on Opus 5. 4

  8. OpenAI released GPT-6 Sol and GPT-6 Luna. 11

  9. OpenAI released GPT-6 Astra, its most capable model. 11

  10. Anthropic launched Claude Fable 5.1 at $10 and $50, with cache reads at $0.25. 4

Questions people ask

Is GPT-4 still available?

In the API, until 23 October 2026. OpenAI deprecated gpt-4 and gpt-4-turbo on 22 April 2026 and named gpt-5.6-sol as the replacement. GPT-4 costs $30 input and $60 output per million tokens with an 8,192-token context window; GPT-6.1 Sol costs $2 and $10 with a 1,050,000-token window, so moving is also cheaper. 121310

Which is cheaper, Claude or GPT?

At list price they match tier for tier: $10 and $50 per million tokens at the top, $2 and $10 in the middle, $0.10 and $0.50 at the bottom. The differences are in the rules. OpenAI charges 2x input above 272K tokens, Claude Haiku 5.5 charges 5x above 100,000 tokens, and Claude Fable 5.1 reads cached input for a quarter of what GPT-6 Astra charges. 210

Is Claude better than GPT for coding?

At the top end, on the two independent rankings read for this page, yes. Artificial Analysis scored Claude Opus 5.5 at 58 and GPT-6 Astra at 53 on an index where coding carries 20% of the weight, and Arena ranks Opus 5.5 first for agent work with Astra third. The Arena rank ranges overlap, and GPT-6.1 Sol gets close for far less per task, so test both on your own repository. 222324

What is the context window of Claude and GPT-6?

Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 each take 1M tokens, which Anthropic puts at roughly 555,000 words. GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna each have a 1,050,000-token window with a 922,000-token input limit. All seven can write up to 128,000 tokens in one reply. 19

Do Claude and ChatGPT train on what I type?

Through the API, not by default on either: Anthropic does not train on inputs or outputs from its commercial products, and OpenAI has not trained on API data since 1 March 2023 unless you opt in. Team and Business workspaces are excluded by default too. On personal plans it is a setting: Claude lists model training as opt-out on Free, Pro and Max, and ChatGPT has an "Improve the model for everyone" switch under Data controls. 61431821

Should I pay for Claude Pro or ChatGPT Plus?

Both are $20 a month. Claude Pro adds the Opus models, Claude Code and research; Fable is only available on Pro through paid usage credits. ChatGPT Plus adds Codex with GPT-6.1 Sol and GPT-6 Luna, and OpenAI also sells a Go plan at $8. Pick Claude Pro to write and code with Claude, and ChatGPT if you want Codex or the lower price. 318

Sources

Each page was read on the date beside it. The links open the current version, which may have changed since.

  1. 1Anthropic: Claude models overview
  2. 2Anthropic: Claude API pricing
  3. 3Anthropic: Claude plans and pricing
  4. 4Anthropic: Claude Platform release notes
  5. 5Anthropic: model deprecations
  6. 6Anthropic: is my data used for model training (commercial products)
  7. 7Anthropic: Claude Code overview
  8. 8OpenAI: API models
  9. 9OpenAI: GPT-6 Astra model page
  10. 10OpenAI: API pricing
  11. 11OpenAI: API changelog
  12. 12OpenAI: API deprecations
  13. 13OpenAI: GPT-4 model page
  14. 14OpenAI: your data in the API
  15. 15OpenAI: gpt-oss-120b model page
  16. 16OpenAI: OpenAI on Amazon Bedrock
  17. 17Microsoft: Foundry Models sold by Azure
  18. 18OpenAI: ChatGPT and Codex pricing
  19. 19OpenAI: ChatGPT and Codex models
  20. 20OpenAI: ChatGPT desktop app
  21. 21OpenAI: data controls FAQ
  22. 22Artificial Analysis: LLM leaderboard
  23. 23Artificial Analysis: Intelligence Index methodology
  24. 24Arena: agent leaderboard

Take the free Starter Kit.

Five working files from the IImagined programs: 32 Instagram hooks, an AI character template, a pricing worksheet and two calculators. Free, and they open in your browser.

Get the free Starter Kit

Related program

AI SaaS Builder. Validate an idea, build on Supabase and Next.js, add AI with the Claude API, use Claude Code and MCP, deploy on Vercel, launch and bill with Stripe. Included in All Access.

IImagined publishes this comparison and sells the All Access membership.