Anthropic Surpasses OpenAI With $965 Bn Valuation, Launches Claude Opus 4.8

Anthropic has raised $65 billion in a Series H funding round, valuing the AI company at $965 billion post-money. The company also launched Claude Opus 4.8, the latest version of its flagship AI model.

Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital led the funding round. Anthropic said the capital will be used to advance safety and interpretability research, expand compute infrastructure, and scale products and partnerships as demand for Claude continues to grow.

The funding round makes Anthropic the world’s most valuable AI startup, surpassing OpenAI’s most recent valuation of $852 billion. The AI company raised $122 billion in March. 

The announcement comes as Anthropic reported that its annualised revenue run rate crossed $47 billion earlier this month, driven by increasing enterprise adoption of Claude.

“Claude is increasingly indispensable to our growing global community of customers, and we work tirelessly to make tools like Claude Code and Cowork more helpful, more powerful, and more adaptable to their needs,” said Krishna Rao, Chief Financial Officer of Anthropic.

“This funding will help us serve the historic demand we are experiencing, stay at the research frontier, and bring Claude to more of the places where work happens,” Rao added.

The round also included $15 billion in previously committed investments from hyperscalers, including $5 billion from Amazon.

Anthropic said strategic infrastructure partners Micron, Samsung, and SK hynix also joined the round. The company said these partnerships would help support future compute requirements as demand for Claude grows.

The company added that it has recently expanded compute capacity through agreements with Amazon for up to five gigawatts of new infrastructure, and with Google and Broadcom for five gigawatts of next-generation TPU capacity. 

Anthropic also secured access to GPU capacity in SpaceX’s Colossus 1 and Colossus 2 systems.

Opus 4.8 With Improved Coding, Agentic Capabilities

Anthropic said Claude Opus 4.8 improves performance across coding, reasoning, agentic tasks and practical knowledge work benchmarks while maintaining the same pricing as Opus 4.7.

The model tops benchmarks like 69.2% on SWE-Bench Pro and 74.6% on Agentic Terminal Coding, beating rivals including OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro.

“Early testers have found Claude Opus 4.8 to be more reliable and sharper in its judgement when it’s performing agentic tasks,” Anthropic said.

Pricing remains unchanged at $5 per million input tokens and $25 per million output tokens for regular usage. Fast mode pricing is set at $10 per million input tokens and $50 per million output tokens.

Claude Opus 4.8 is available through claude.ai and the Claude API under the model name claude-opus-4-8.

The company also launched dynamic workflows for Claude Code in research preview. The feature enables Claude to plan tasks and run hundreds of parallel subagents within a single session, allowing it to execute large-scale code migrations and verify outputs before delivering results.

Anthropic further introduced effort controls on claude.ai and Cowork, giving users the ability to adjust how much compute the model uses for a task. Higher effort settings provide deeper reasoning and longer processing, while lower settings prioritise speed and lower rate-limit consumption.

According to the company, Opus 4.8 is around four times less likely than Opus 4.7 to leave flaws in generated code unreported.

“One of the most prominent improvements in Opus 4.8 is its honesty,” Anthropic said. “Opus 4.8 is more likely to flag uncertainties about its work and less likely to make unsupported claims.”

Besides that, Anthropic said it plans to make Mythos-class models more widely available after introducing additional cybersecurity safeguards.

The company also updated its Messages API, allowing developers to insert system entries directly into message arrays to modify instructions during a task without interrupting prompt caching.