Claude Beats ChatGPT on Coding as OpenAI Holds Its Reach

Terminal-Bench 4.0 gives Claude a clear lead over GPT-6 Astra: 66.4% to 57.9%. The gap narrows on other tests, and the results do not settle which service will work better on a particular team's codebase.

Anthropic released Claude Opus 5.5 on September 22, 2026. OpenAI's GPT-6.1 Sol arrived on September 29, while GPT-6 Luna remains positioned for high-volume document tasks.

Anthropic followed with Sonnet 5.5 on September 28, then released Haiku 5.5 on October 7. Both lineups now include models with context windows of about one million tokens.

Where the benchmark results split

Claude also led Toolathlon, scoring 61.7% against GPT-6 Astra's 58.0%. In the cited prompt-injection test, Claude recorded a lower attack success rate: 4.8%, compared with 8.5% for GPT-6 Astra. That points to an advantage on practical, multi-step work. It does not show that either model is immune to attacks or will perform the same way on every codebase.

OpenAI's flagship GPT-6 Astra led the listed FrontierMath Tier 4 comparison, scoring 97.6% against 87.8% for Claude Fable 5. GPT-6 Astra also scored 96.0% on GPQA Diamond. On GDPval-AA, Claude scored 1,846 against ChatGPT's 1,542. The cited SWE-Bench Pro results put Claude Fable 5 at 80.0%, while GPT-6 landed in a comparable mid-70s to low-80s range.

The tests cover different tasks, and some compare different models. Teams choosing an API should run both services on the same codebase with a fixed token budget. Dependencies and coding conventions can shift the result; so can the team's test framework.

API costs and Australian subscriptions

Claude Pro and ChatGPT Plus each cost about US$20 a month at the standard paid consumer tier. OpenAI lists ChatGPT Plus at US$20 per month and includes image generation. Its free plan offers limited image generation. Voice access is also limited on the free plan, with expanded limits on paid plans, as outlined on the ChatGPT pricing page.

The comparison estimates Australian prices at roughly $32 monthly for either plan. ChatGPT Pro is about $320, while Claude Max ranges from $160 to $320. AUD prices vary with exchange rates, and local checkout prices can reflect currency conversion and taxes.

For API users, Claude Opus 5.5 is listed at US$4 per million input tokens and US$20 per million output tokens. The comparison puts a GPT-6 Astra-class endpoint at US$10 and US$50 respectively. Claude Opus also lists cached input at US$0.20 per million tokens, against its standard US$4 rate. That lower rate may cut costs on repeated prompts and reference material.

The mid-tier comparison is closer on price. Claude Sonnet 5.5 and GPT-6.1 Sol are both listed at US$2 per million input tokens and US$10 per million output tokens. For bulk work, Claude Haiku 5.5 and GPT-6 Luna each list US$0.10 input and US$0.50 output per million tokens under the stated threshold. Haiku's rates rise above 100,000-token prompts to US$0.50 input and US$2.50 output per million tokens.

OpenAI's wider business push also reaches financial services. Its offering pairs GPT-6 Astra with live market data for investment-bank research and modelling, as described in this financial services report.

Long documents versus an all-in-one assistant

Both services offer million-token context windows across several models, so context size alone no longer separates them at the top end. Claude is geared toward retrieving details from large document collections. It can also produce long-form writing that follows detailed style instructions. ChatGPT's browsing and deep-research tools can gather material from multiple online sources, while Claude can help synthesise documents already collected.

ChatGPT combines voice with image generation. It also offers web browsing. OpenAI says Voice lets users speak with ChatGPT and hear spoken responses. In Live conversations, speech and streamed text appear together, and the chat can use web search and memory.

OpenAI's Voice guidance lists up to three hours a day for Plus. The US$100 Pro plan allows up to 15 hours. The US$200 Pro plan includes unlimited use of GPT-Live-1, while free access is limited. Availability and limits can vary by plan and account settings, as detailed in the ChatGPT Voice guide.

ChatGPT can generate images in chat. As of October 2026, Claude has no native voice mode or built-in image generator. For a marketing team creating copy and graphics, that gap may matter more than a coding benchmark. It may matter less to teams handling long documents.

Market-share figures cited from First Page Sage put ChatGPT at 52.1% in September 2026. Claude stood at 10.6%. Google Gemini had 26.9%.

Claude's consumer share is smaller, but a FatJoe statistics roundup reported that 43.8% of surveyed US businesses paid for Anthropic services in August, compared with 39.8% for OpenAI. Anthropic reportedly has more than 300,000 business customers, including eight of the Fortune 10.

For Australian organisations, both providers offer business tiers with retention controls. They also offer options to exclude conversations from model training. Neither publishes a dedicated Australian data-residency option as of October 2026. Organisations handling client or patient data need to check processing locations against their requirements. They should also review training settings and contract terms.