# Claude vs ChatGPT

> Source: https://aiwiki.ai/wiki/claude_vs_chatgpt
> Updated: 2026-07-07
> Categories: AI Models, Conversational AI, Large Language Models
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution to "AI Wiki (aiwiki.ai)".

As of July 2026, neither [Claude](/wiki/claude) nor [ChatGPT](/wiki/chatgpt) is universally better: the honest answer depends on the task. Claude, from [Anthropic](/wiki/anthropic), leads on agentic coding, long-horizon reasoning, and enterprise reliability, while ChatGPT, from [OpenAI](/wiki/openai), wins on multimodal breadth (native image, voice, and video generation), consumer reach, and price at the mid tier. If your work is code, technical writing, or long-document analysis, Claude's current flagships (Claude Fable 5, [Claude Opus 4.8](/wiki/claude_opus_4_8), and Claude Sonnet 5) are the stronger default. If you want one app that also makes images, talks back in real time, and generates video for a low monthly price, ChatGPT (running GPT-5.5) is the more complete package.

By the numbers: Claude Fable 5 tops the [Artificial Analysis](/wiki/artificial_analysis) Intelligence Index at 60, ahead of Claude Opus 4.8 at 56 and GPT-5.5 at 55 [7]. On the harder [SWE-Bench Pro](/wiki/swe_bench_pro) coding test Claude leads decisively (Fable 5 80.3%, Opus 4.8 69.2%, versus GPT-5.5 at 58.6%) [8], while the two labs are within a tenth of a point on the near-saturated [SWE-bench](/wiki/swe_bench) Verified split (Opus 4.8 88.6%, GPT-5.5 88.7%) [14][5]. Both families now ship a 1M-token [context window](/wiki/context_window) at the top tier [1][7].

## Capability-by-task: which wins where?

| Task | Claude edge | ChatGPT edge | Winner |
| --- | --- | --- | --- |
| Agentic / repo-scale coding | Fable 5, Opus 4.8 top SWE-Bench Pro [8] | GPT-5.5 competitive on SWE-bench Verified [5] | Claude |
| Everyday coding assistant | Sonnet 5 strong, 1M context [1] | GPT-5.4 cheaper per token [4] | Toss-up |
| Hard science / reasoning | AA Index leader [7] | [GPQA](/wiki/gpqa) Diamond 93.6% [5] | Toss-up |
| Long-document analysis | 1M context, high recall [1] | 1M context on GPT-5.5 [7] | Tie |
| Professional writing | Preferred for tone, editing | Strong, more templated | Claude (slight) |
| Deep research / web browsing | Research mode (Pro) [3] | Broader tools, Deep Research [9] | ChatGPT (slight) |
| Image generation | None (input only) [1] | Native [gpt-image-1](/wiki/gpt_image) [11] | ChatGPT |
| Voice conversation | Beta voice, 18 languages [10] | Mature [Advanced Voice Mode](/wiki/advanced_voice_mode) [9] | ChatGPT |
| Video generation | None | [Sora](/wiki/sora) built in [11] | ChatGPT |
| Enterprise deployment | 80% of revenue is business [13] | Large enterprise tier too [9] | Claude (slight) |

Last verified: July 2026.

## Head-to-head spec and price table

The brand-level comparison across the dimensions people ask about most. Prices are USD; API prices are per 1,000,000 tokens.

| Spec | Claude (Anthropic) | ChatGPT (OpenAI) |
| --- | --- | --- |
| Top model, generally available | Claude Fable 5 [2] | GPT-5.5 / GPT-5.5 Pro [4][5] |
| Coding, SWE-bench Verified | 95.0% (Fable 5); 88.6% (Opus 4.8) [14] | 88.7% (GPT-5.5) [5] |
| Coding, SWE-Bench Pro (harder) | 80.3% (Fable 5); 69.2% (Opus 4.8) [8] | 58.6% (GPT-5.5) [8] |
| Reasoning, GPQA Diamond | n/r; retired as saturated (Opus 4.8 approx 92.4% per third party) [15] | 93.6% (GPT-5.5) [5] |
| Reasoning, AA Intelligence Index | 60 (Fable 5); 56 (Opus 4.8) [7] | 55 (GPT-5.5) [7] |
| Context window | 1M tokens [1] | 1M tokens (GPT-5.5) [7] |
| Image generation | No, text and image input only [1] | Yes, native gpt-image-1 [11] |
| Voice | Beta voice mode, mobile, 18 languages [10] | Advanced Voice Mode, real time [9] |
| Video generation | No | Sora, built into paid plans [11] |
| Consumer plan price | Pro 17 to 20/mo; Max from 100/mo [3] | Plus 20/mo; Go 8/mo; Pro 100 or 200/mo [9] |
| API price, flagship (in / out) | 10 / 50 (Fable 5); 5 / 25 (Opus 4.8) [1] | 5 / 30 (GPT-5.5); 30 / 180 (5.5 Pro) [4] |
| Weekly active users | Consumer base smaller; enterprise-led [13] | Roughly 900M weekly [12] |

Last verified: July 2026.

## Current flagship models compared

ChatGPT's underlying models have advanced from [GPT-4](/wiki/gpt_4) and [GPT-4o](/wiki/gpt_4o) through the [GPT-5](/wiki/gpt_5) generation to today's GPT-5.5, while Claude has moved from Opus 4.x to the new Fable tier. All models below are proprietary. Claude prices are from Anthropic's model docs; GPT prices are from OpenAI's API pricing page. AA = Artificial Analysis Intelligence Index.

| Model | Developer | Release | Context | SWE-bench Verified | SWE-Bench Pro | GPQA Diamond | AA Index | Input $/1M | Output $/1M |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Claude Fable 5 | Anthropic | 2026-06 | 1M | 95.0% [14] | 80.3% [8] | retired [15] | 60 [7] | 10 [2] | 50 [2] |
| [Claude Opus 4.8](/wiki/claude_opus_4_8) | Anthropic | 2026 | 1M | 88.6% [14] | 69.2% [8] | 92.4% [15] | 56 [7] | 5 [1] | 25 [1] |
| Claude Sonnet 5 | Anthropic | 2026-06 | 1M | approx 80.9% [15] | 63.2% [8] | retired [15] | n/r | 3 [1] | 15 [1] |
| GPT-5.5 | OpenAI | 2026-04 | 1M | 88.7% [5] | 58.6% [8] | 93.6% [5] | 55 [7] | 5 [4] | 30 [4] |
| GPT-5.5 Pro | OpenAI | 2026-04 | 1M | n/r | n/r | n/r | n/r | 30 [4] | 180 [4] |

Last verified: July 2026. Blank cells (n/r) are values the provider does not report. Claude Sonnet 5 carries introductory API pricing of 2 / 10 through August 31, 2026 [1]. A restricted top tier exists on each side but is not generally available: Anthropic's Claude Mythos 5 (invitation-only, Project Glasswing) [2] and OpenAI's GPT-5.6 Sol, Terra, and Luna (limited preview announced June 26, 2026) [6].

## Which is better for coding?

Claude is the stronger coder in 2026, and the gap widens on harder tests. On the near-saturated SWE-bench Verified split, GPT-5.5 (88.7%) and Claude Opus 4.8 (88.6%) are effectively tied, with Claude Fable 5 pulling ahead at 95.0% [5][14]. Both labs caution that Verified is close to saturated and that frontier models may have trained on adjacent data, so raw Verified scores should be read skeptically [14]. The more discriminating benchmark is SWE-Bench Pro, which uses harder, less-contaminated tasks: there Claude leads clearly, with Fable 5 at 80.3%, Opus 4.8 at 69.2%, and Sonnet 5 at 63.2%, versus GPT-5.5 at 58.6% [8]. On the independent, neutral-harness [Terminal-Bench](/wiki/terminal_bench) leaderboard the order narrows, with GPT-5.5 at 78.2% ahead of Opus 4.8 at 74.6% [16]. In practice, Claude powers dedicated coding surfaces like [Claude Code](/wiki/claude_code) and [Claude Cowork](/wiki/claude_cowork), while OpenAI ships [Codex](/wiki/codex); teams doing large, multi-file refactors tend to prefer Claude, while ChatGPT remains a capable, cheaper generalist for everyday snippets.

## Which is better at reasoning and knowledge?

This one is close to a tie at the frontier. On GPQA Diamond, a graduate-level science exam, GPT-5.5 scores 93.6% [5], and Claude Opus 4.8 sits around 92.4% per third-party testing [15]. Anthropic has stopped reporting GPQA Diamond and [AIME](/wiki/aime) for its newest models, declaring both saturated after scores crossed roughly 94% and 99% respectively, so Fable 5 and Sonnet 5 publish no official GPQA number [15]. The better cross-model signal is now the Artificial Analysis Intelligence Index, a composite of nine harder evaluations including [Humanity's Last Exam](/wiki/humanitys_last_exam), [FrontierMath](/wiki/frontiermath)-style problems, and Terminal-Bench. There Claude Fable 5 leads at 60, Opus 4.8 at 56, and GPT-5.5 at 55 [7]. Net: Claude has a narrow lead on the hardest aggregate reasoning tests, while ChatGPT matches or edges it on classic single-benchmark science questions.

## Which has the bigger context window?

They are tied at the top. Claude Fable 5, Opus 4.8, and Sonnet 5 each offer a 1M-token context window (roughly 555,000 words) with up to 128k tokens of output [1]. GPT-5.5 also advertises a 1M-token window, though Artificial Analysis measured an effective length closer to 922k tokens [7]. For most users the practical difference is negligible; both can hold a large codebase or a long set of documents in a single prompt. Claude's smaller, faster [Claude Haiku 4.5](/wiki/claude_haiku_4_5) keeps a 200k window [1], and OpenAI's cheaper GPT-5.4 tier is the value option on the ChatGPT side [4].

## Which is better for images, voice, and video?

ChatGPT wins multimodal generation outright. ChatGPT generates images natively through gpt-image-1 (the successor to [DALL-E](/wiki/dall_e)), produces video through Sora, and offers a mature, real-time Advanced Voice Mode [11][9]. Claude, by contrast, accepts image input and has excellent vision, but it cannot generate images or video: its models are text-output only [1]. Claude does have a conversational voice mode, which exited beta for multilingual input in June 2026 and now supports 18 languages on the mobile apps [10]. So for any workflow that needs pictures, audio, or clips out of the box, ChatGPT is the clear choice; for pure text and code, the lack of image generation rarely matters.

## Which is cheaper?

It depends on tier. For consumers, ChatGPT is cheaper at the entry level: a Go plan runs 8/mo and Plus is 20/mo, versus Claude Pro at 17/mo on annual billing or 20/mo month to month [9][3]. Both offer high-usage tiers around 100 to 200/mo (Claude Max, ChatGPT Pro) [3][9]. On the API, GPT-5.5 at 5 input and 30 output per million tokens is close to Claude Opus 4.8 at 5 / 25, while Claude Sonnet 5 (3 / 15, or 2 / 10 introductory) and OpenAI's GPT-5.4 (2.50 / 15) are the mid-tier value picks [4][1]. Claude Fable 5 is the most expensive flagship at 10 / 50 [2]. Note that Claude's newer tokenizer can inflate token counts by roughly 1.0 to 1.35x versus older models, which narrows Sonnet 5's headline price advantage in real use [1].

## Which has the better app ecosystem and reach?

ChatGPT is the consumer giant; Claude is the enterprise and developer favorite. OpenAI reported roughly 900 million weekly active ChatGPT users in early 2026 and around 1 billion monthly users by mid-2026, alongside apps for web, mobile, and desktop, the [ChatGPT Atlas](/wiki/chatgpt_atlas) browser, custom GPTs, Sora, and Codex [12][9]. Anthropic runs a smaller consumer footprint but derives about 80% of revenue from business, with 300,000-plus business customers and a run-rate reported near 14 billion in early 2026 under CEO [Dario Amodei](/wiki/dario_amodei) [13]. Claude's ecosystem leans developer-first: Claude Code, Claude Cowork, and the [Model Context Protocol](/wiki/model_context_protocol) standard that Anthropic created and OpenAI later adopted. Both trail [Gemini 3](/wiki/gemini_3) on some raw benchmarks but lead it on product breadth. If you want the widest set of consumer features in one app, pick ChatGPT; if you are building on an API or standardizing an engineering org, Claude is the more common enterprise default.

## Which should you choose?

Use-case verdicts, based on the July 2026 evidence above:

- Coding and software engineering: Choose Claude. It leads SWE-Bench Pro by a wide margin (Fable 5 80.3%, Opus 4.8 69.2% versus GPT-5.5 58.6%) and ships purpose-built tools in Claude Code and Cowork [8].
- Writing and editing: Slight edge to Claude for tone and revision control; ChatGPT is close and cheaper.
- Research, images, voice, and video: Choose ChatGPT. It is the only one of the two that generates images and video and has the more mature voice mode [11][9].
- Everyday all-rounder on a budget: Choose ChatGPT (Go at 8/mo or Plus at 20/mo) [9].
- Enterprise, agents, and long-context reliability: Choose Claude, which is engineered around agentic and enterprise workloads and dominates business revenue [13].
- Absolute maximum reasoning: Choose Claude Fable 5 (AA Index 60), with GPT-5.5 a close second [7].

Both are excellent [large language model](/wiki/large_language_model) assistants; the right pick is the one whose strengths match your primary workload. Many teams run both: Claude for code and agents, ChatGPT for consumer-facing, multimodal, and research tasks.

## References

1. Anthropic, "Models overview," Claude Platform Docs, platform.claude.com/docs/en/about-claude/models/overview (accessed July 2026). Specs, context windows, and API pricing for Claude Fable 5, Opus 4.8, Sonnet 5, and Haiku 4.5.
2. Anthropic, "Introducing Claude Fable 5 and Claude Mythos 5," Claude Platform Docs (June 2026). Fable 5 general availability June 9, 2026; pricing 10 / 50 per 1M tokens; safety classifiers.
3. Anthropic, "Pricing," claude.com/pricing (accessed July 2026). Free, Pro (17 to 20/mo), Max (from 100/mo) consumer plans.
4. OpenAI, "API Pricing," developers.openai.com/api/docs/pricing (accessed July 2026). GPT-5.5 (5 / 30), GPT-5.5 Pro (30 / 180), GPT-5.4 tiers.
5. OpenAI, "Introducing GPT-5.5," openai.com/index/introducing-gpt-5-5 (April 23, 2026). SWE-bench Verified 88.7%, GPQA Diamond 93.6%, Terminal-Bench results.
6. OpenAI, "Previewing GPT-5.6 Sol," openai.com/index/previewing-gpt-5-6-sol (June 26, 2026). Sol, Terra, Luna limited preview.
7. Artificial Analysis, Intelligence Index and model pages, artificialanalysis.ai (accessed July 2026). Index scores: Fable 5 60, Opus 4.8 56, GPT-5.5 55; context measurements.
8. SWE-Bench Pro leaderboard and Fable 5 comparison, morphllm.com/swe-bench-pro (accessed July 2026). Fable 5 80.3%, Opus 4.8 69.2%, Sonnet 5 63.2%, GPT-5.5 58.6%.
9. OpenAI, "ChatGPT Plans and Pricing," chatgpt.com/pricing (accessed July 2026). Free, Go (8/mo), Plus (20/mo), Pro (100 and 200/mo), Business, Enterprise.
10. Anthropic, "Using voice mode on Claude Mobile Apps," support.claude.com (updated June 2026). Beta voice mode, 18 languages.
11. OpenAI, image generation (gpt-image-1) and Sora documentation, openai.com (accessed July 2026). Native image and video generation in ChatGPT.
12. OpenAI weekly active user disclosures and Reuters reporting (Feb to June 2026). Roughly 900M weekly and 1B monthly ChatGPT users.
13. Anthropic revenue and enterprise statistics, Anthropic and DemandSage (2026). About 80% business revenue, 300,000-plus business customers, run-rate near 14 billion.
14. Claude benchmarks aggregation, morphllm.com/claude-benchmarks (accessed July 2026). SWE-bench Verified: Fable 5 95.0%, Opus 4.8 88.6%; saturation and memorization caveats.
15. "Best AI Models July 2026," felloai.com/best-ai-models and aivy.com.au Fable 5 review (2026). Opus 4.8 GPQA Diamond approx 92.4%; Anthropic retiring GPQA/AIME as saturated; Sonnet 5 figures.
16. "How OpenAI's GPT-5.5 stacks up," rdworldonline.com (2026). Independent Terminal-Bench neutral-harness scores: GPT-5.5 78.2%, Opus 4.8 74.6%.