Anthropic just launched Claude Opus 4.8 and became the world's most valuable AI startup in the same week. On Thursday, the company closed a $65 billion funding round that values it at $900 billion, ahead of OpenAI's last valuation of $730 billion.
If you're a developer or running a tech business, this is worth your attention.
Claude Opus 4.8 Targets the AI Reliability Problem
The biggest complaint about AI tools in professional settings, especially among fintech and legal teams adopting AI cautiously, is overconfidence. Models confidently give wrong answers. That costs time and trust.
Anthropic says Opus 4.8 addresses this directly. Testers who used the model early say it hesitates where previous models would have pushed forward. Bridgewater's Michael Ran, who joined an early test group, said the model caught analysis errors that previous tools left for humans to find.
For fintech teams or legal analysts running AI on real decisions, that's the gap that gets someone fired.
Opus 4.8 scored 10 per cent higher than its predecessor on the Vals AI vibecoding benchmark. That test measures how well models write software from plain-English prompts.
READ: AI Is Creating More Cybersecurity Jobs Than It Kills
Who Put In the $65 Billion
The round was led by Greenoaks Capital, Sequoia, Altimeter, and Dragoneer — with Samsung, Micron, and SK Hynix joining as chip-sector strategic partners. Amazon and Google were already on the cap table.
Anthropic says its annualised revenue has now crossed $47 billion. Its valuation has grown nearly two and a half times from $380 billion just three months ago.
OpenAI needed a decade to hit its current valuation. Anthropic, founded in 2021, got there in under five years.
What "Mythos-Class Models" Actually Means
Anthropic did not just announce a new model. It teased a new category.
The company says it is testing "Mythos-class models" for public release in the coming weeks. " Mythos is the internal model Anthropic has described as capable of finding and exploiting hidden flaws in software powerful enough that it has not yet been released publicly. Cybersecurity concerns have delayed the rollout.
Whether any of that capability survives into a commercial product is another question. AI companies routinely build expectations that reality then deflates. But if even a fraction of Mythos's reported capability ships in a commercial model, it would matter to security teams already stretched thin.
READ: ClickUp Cuts 22% of Staff and Offers $1M Pay for AI-Powered Workers
Two Features Worth Knowing Right Now
Anthropic also shipped two features alongside Opus 4.8 that address real workflow friction.
Dynamic workflows, now in research preview, let Claude deploy hundreds of subagents in parallel for complex coding tasks. Think of it as offloading a project to a team, not a single assistant.
Effort control adds a manual slider to Claude's model selector. You pick low, medium, high, or max depending on how much computing power you want the model to use. Adaptive thinking mode is still available as a toggle.
The previous model, Opus 4.7, drew complaints because its adaptive thinking feature misjudged task difficulty too often. This is Anthropic's direct response.
