OpenAI, Gemini and Claude: where the big three AI providers stand now

September 2026 was an unusually busy month for the three companies behind the most widely used AI assistants. OpenAI released GPT-6 Astra on 3 September and two cheaper GPT-6 models later in the month. Anthropic launched Claude Fable 5.1 on 1 September, then Claude Opus 5.5 and Claude Sonnet 5.5 in the last week. Google made Gemini 3.8 Flash generally available on 2 September and, on 30 September, announced Gemini 4 Argon, which is so far available only to a small group of security specialists. This article sets out where each company stands as of 5 October 2026, asking the same questions of all three and using each company's own documentation, two independent leaderboards and public financial reporting. NerdBible has not run its own tests of these models; every figure below links to where it comes from, and prices are the US prices on the official pages on that date.

OpenAI: ChatGPT and the OpenAI API

Models

OpenAI's developer documentation names three current flagship models: GPT-6 Astra, "our flagship model for complex reasoning and coding"; GPT-6.1 Sol, to "balance intelligence and cost"; and GPT-6 Luna for "cost-sensitive, high-volume workloads." All three accept text and image input, produce text, list a context window of 1.05 million tokens, and support function calling, web search, file search and computer use (OpenAI models). The API changelog dates GPT-6 Astra to 3 September, GPT-6 Sol and GPT-6 Luna to 22 September and GPT-6.1 Sol to 29 September (OpenAI changelog). OpenAI also lists separate models for voice conversations, transcription and image generation, plus cybersecurity and life sciences models offered to approved organisations.

API pricing

At standard rates for prompts in the short-context band, GPT-6 Astra costs $10 per million input tokens and $50 per million output tokens; GPT-6.1 Sol $2 and $10; GPT-6 Luna $0.10 and $0.50. Longer prompts cost more, and Batch and Flex processing halve the standard rates (OpenAI pricing).

ChatGPT plans

The ChatGPT pricing page lists Free, Go, Plus and Pro for individuals. Free includes "Unlimited text chats with GPT-5.6 Luna," subject to abuse guardrails, with limited uploads, image creation, voice chats, deep research and Codex access. OpenAI notes that the Go plan "may include ads." Plus adds GPT-6 models including GPT-6 Astra, expanded deep research and more Codex use (ChatGPT pricing). OpenAI's help centre puts Plus at $20 a month (OpenAI Help). Pro comes in $100 and $200 tiers; OpenAI says that since 10 September it has temporarily paused new sign-ups to the $200 tier (OpenAI Help). For organisations, ChatGPT Business standard seats cost $25 per user per month, or $20 billed annually, with Premium seats at $125, or $100 annually (OpenAI Help).

Data use

For individual services, OpenAI says "we may use your content to train our models," and that you can opt out by turning off "Improve the model for everyone" in Data controls. Temporary chats are not used for training. By default, OpenAI does not use inputs or outputs from ChatGPT Business, Enterprise, Edu or the API to improve its models (OpenAI Help).

Clouds and funding

OpenAI documents its models on Amazon Bedrock, including GPT-6 Astra (OpenAI on Amazon Bedrock), and Microsoft lists GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna among the models sold directly by Azure in Microsoft Foundry (Microsoft Learn). CNBC reported that OpenAI closed a $122 billion funding round at an $852 billion post-money valuation in March, with Amazon agreeing to invest up to $50 billion and Nvidia and SoftBank $30 billion each (CNBC).

Google: the Gemini app and the Gemini API

Models

Google's line-up is split between what is generally available and what has only been announced. On 30 September Google announced Gemini 4 Argon as "our new frontier model," rolling out first to "a set of trusted cyber defenders" through its Fairwind Program, with wider availability to developers, enterprises and consumers to follow. Google says it will launch at an introductory API price of $2 per million input tokens and $10 per million output tokens (Google). Among models anyone can use, the Gemini app describes Gemini 3.1 Pro as "our leading model" (Gemini subscriptions). It was released in preview on 19 February, and Gemini 3.8 Flash, "our most intelligent Flash model," became generally available on 2 September (Gemini API release notes). Google also offers Flash-Lite models, image generation (Nano Banana), video generation (Gemini Omni Flash), music and speech models (Gemini models).

API pricing and context

Gemini 3.1 Pro Preview costs $2 per million input tokens and $12 per million output tokens for prompts up to 200,000 tokens, and $4 and $18 above that. Gemini 3.8 Flash costs $0.75 and $3.75 through 31 December 2026, rising to $1.50 and $7.50 from 1 January 2027. The free tier of the API is marked "Used to improve our products: Yes," the paid tier "No" (Gemini API pricing). Both Gemini 3.1 Pro and Gemini 3.8 Flash accept 1,048,576 input tokens (Gemini 3.1 Pro model page).

Gemini app plans

In the US, the free plan includes Gemini 3.6 Flash, "varying access to 3.1 Pro," image generation, Deep Research, Gemini Live and Canvas. Google AI Plus costs $4.99 a month, Google AI Pro $19.99 a month with 4x the free usage and Gemini inside Gmail and Docs, and Google AI Ultra starts at $99.99, with a $199.99 tier. The paid plans bundle Google storage and other services (Gemini subscriptions).

Data use

Google says that if you are 18 or over, the Keep Activity setting "is on by default" (Gemini Apps Help). With it on, Google uses your activity "to provide, develop, and improve its services (including training generative AI models)," with help from human reviewers; reviewed data is disconnected from your account and kept for three years. Temporary chats and chats with Keep Activity off are kept for 72 hours and "are not used to train Google's AI models" (Gemini Apps Privacy Hub).

Clouds and business

Gemini models are offered to businesses through Google Cloud's Gemini Enterprise Agent Platform, which lists Gemini 3.1 Pro and the Flash models (Google Cloud). In its second-quarter results, Alphabet said the Gemini app had 950 million monthly active users and that "nearly 90% of the Fortune 100" were using Gemini Enterprise (Alphabet Q2 results).

Anthropic: Claude and the Claude API

Models

Anthropic's documentation lists four current models: Claude Fable 5.1 "for demanding reasoning and long-horizon agentic work," Claude Opus 5.5 "for long-running agentic coding and knowledge work," Claude Sonnet 5.5, "the best combination of speed and intelligence," and Claude Haiku 4.5, the fastest. It advises starting with Opus 5.5 for most workloads. The first three have a 1 million token context window; Haiku 4.5 has 200,000. All accept text and image input and produce text (Claude models overview). The release notes date Fable 5.1 to 1 September, Opus 5.5 to 22 September and Sonnet 5.5 to 28 September. Like OpenAI and Google, Anthropic also has a restricted model, Claude Mythos 5.1, available to participants in its Project Glasswing programme (Claude release notes).

API pricing

Per million input and output tokens: Fable 5.1 $10 and $50, Opus 5.5 $4 and $20, Sonnet 5.5 $2 and $10, Haiku 4.5 $1 and $5 (Claude models overview).

Claude plans

Claude Free includes chat on web, desktop and mobile, web search, file creation, memory, voice mode and Sonnet and Haiku models. Pro costs $20 a month, or $17 a month billed annually, and adds Opus, Claude Code and Claude in Chrome; Max starts at $100 a month with 5x or 20x Pro's usage. Fable is not on the Free plan. Team standard seats cost $25 per seat per month, or $20 billed annually, and Premium seats $125, or $100 annually; Enterprise is a seat fee plus usage at API rates (Claude pricing).

Data use

For Free, Pro and Max, Anthropic says it uses chats and coding sessions to improve its models if "you choose to allow us," if a conversation is flagged for safety review, or if you opt into a programme such as its Trusted Tester Program; incognito chats are not used (Anthropic Privacy Center). New users pick the setting at sign-up, and allowing training extends data retention to five years from 30 days. These consumer terms do not apply to Team, Enterprise or API use, including through Amazon Bedrock and Google Cloud (Anthropic). Anthropic's pricing table labels model training on its consumer plans "Opt-out" and lists "No model training on your content by default" for Team.

Clouds and funding

Anthropic lists each current model on Amazon Bedrock, Google Cloud, Microsoft Foundry and its own Claude Platform on AWS (Claude models overview). In May it announced a $65 billion Series H at a $965 billion post-money valuation and said its run-rate revenue had passed $47 billion (Anthropic).

What independent leaderboards show

Two widely cited independent rankings disagree at the top, which is a useful reminder that "best model" depends on what is measured. On LMArena's Text Arena, which ranks models by people's votes in blind side-by-side comparisons, the leaderboard dated 2 October placed Gemini 4 Argon (high) first with a score marked "Preliminary," followed by several Claude models; GPT-6 Astra and GPT-6.1 Sol sat lower in the top 30 (LMArena). On the Artificial Analysis Intelligence Index, which combines a set of benchmark evaluations, Claude Opus 5.5 at its highest setting scored 58, and GPT-6 Astra (max), Gemini 4 Argon (high) and Claude Fable 5.1 (max) each scored 53 (Artificial Analysis). Both boards change often and new models move quickly, so treat any single position as a snapshot.

Side by side

How to choose

• Everyday writing and questions: all three free tiers handle this. The practical differences are usage limits and which models the free plan includes, so try the same task in each before paying.
• Coding: each company now sells a coding agent with its paid plans: Codex with ChatGPT, Claude Code with Claude Pro and above, and Jules and Antigravity with Google AI Pro. If you already work in one of these tools, that is usually the deciding factor.
• Research with sources: ChatGPT and the Gemini app both include deep research on their free plans with limits; Claude's Research feature starts on Pro.
• Working inside Google or Microsoft tools: Google AI Pro puts Gemini into Gmail and Docs. Claude Pro includes Claude for Microsoft 365 and Outlook. OpenAI's models are sold through Microsoft's own Foundry platform for businesses building on Azure.
• Privacy-sensitive work: on all three consumer services, check the training setting yourself, since two of the three use chats by default and the third asks at sign-up. Temporary or incognito chats are excluded from training at all three. For confidential business material, the business plans and paid APIs from all three are not used for training by default.
• Cost at scale: on list prices, Gemini 3.1 Pro and the announced Gemini 4 Argon price are the lowest per token among the three companies' top models, while OpenAI's GPT-6.1 Sol and Anthropic's Sonnet 5.5 are each priced at $2 and $10 per million tokens.

None of these companies is standing still: all three released major models within the last five weeks, and each keeps some models restricted to vetted organisations. Prices, plans and defaults change often, so check the linked pages before you commit.