News

Anthropic expands Claude's Memory Fivefold as Coding Competition Intensifies

• By John K. Waters
• 08/12/2025

Anthropic expanded the context window of its Claude Sonnet 4 AI model to 1 million tokens on Tuesday, a fivefold increase that allows the system to process entire software codebases or dozens of research papers in a single request.

The upgrade comes as artificial intelligence companies compete intensely for enterprise customers, particularly in the lucrative AI coding market where Anthropic has built a strong position against rivals, including OpenAI and Google.

The expanded context window - essentially the AI model's working memory - can handle up to 75,000 lines of code, compared with the previous limit of around 15,000 lines. For text processing, the system can now analyze roughly 750,000 words, more than the entire "Lord of the Rings" trilogy.

"This is really cool because it's one of the big barriers I've seen with customers," Brad Abrams, Anthropic's product lead for Claude, told The Verge. "They have to break up their problems into these small chunks with our existing context window, and with a million tokens, the model can handle the entire scope of the context."

Enterprise focus drives growth
Anthropic has differentiated itself from competitors by focusing primarily on enterprise customers through API sales rather than consumer subscriptions. The company has secured partnerships with major AI coding platforms, including Microsoft's GitHub Copilot, Windsurf, and Anysphere's Cursor.

However, OpenAI's release of GPT-5 last week poses a competitive threat. Cursor, one of Anthropic's key customers, recently switched to using GPT-5 as its default model for new users, with Cursor CEO Michael Truell even participating in OpenAI's product launch.

Pricing adjusts for computational demands
The expanded capability comes with higher costs. Anthropic doubled its pricing for prompts exceeding 200,000 tokens, charging $6 per million input tokens compared with $3 for smaller prompts. Output tokens cost $22.50 per million tokens for large prompts, up from $15.

The company said customers can reduce costs through prompt caching and batch processing, which can cut expenses by up to 50%. OpenAI's GPT-5 offers a 400,000-token context window at $1.25 per million input tokens, making it significantly cheaper than Anthropic's offering for large-scale processing tasks.

Market positioning and customer adoption
Early customers highlighted the practical benefits of the expanded context window. Eric Simons, CEO of Bolt.new, which integrates Claude into browser-based development platforms, said the upgrade allows "developers to work on significantly larger projects while maintaining the high accuracy we need for real-world coding."

Sean Ward, CEO of London-based iGent AI, said the expansion enabled "true production-scale engineering" with multi-day sessions on real-world codebases. The feature is available in public beta for Anthropic API customers with higher-tier access limits and through Amazon Web Services' Bedrock platform. Google Cloud integration is planned but not yet available.

Broader AI competition context
The context window expansion represents the latest move in an escalating competition among AI companies to capture enterprise customers. Larger context windows generally improve AI performance on complex tasks, particularly in software development, where models benefit from seeing entire project structures rather than code fragments.

Google's Gemini models offer context windows up to 2 million tokens, though at significantly higher prices. Meta and other competitors have also expanded their models' memory capabilities in recent months.

Anthropic's announcement follows a week of significant AI industry activity, including OpenAI's release of GPT-5 and the company's first open-source models since 2019. The rapid succession of product releases reflects intensifying competition as companies seek to establish market leadership in enterprise AI applications.

The Claude Sonnet 4 upgrade is available immediately for qualifying customers, with broader availability planned over the coming weeks. Anthropic said it is exploring how to bring expanded context capabilities to other Claude products beyond the API.