Anthropic’s fan-favorite coding model, Claude Sonnet 4, just leveled up to a 1 million token context window for enterprise customers.
That’s… insane.
We’re talking 750,000 words or 75,000 lines of code in a single prompt. FYI: that’s enough to fit the entire “Lord of the Rings” trilogy — plus some bonus fanfiction — into its brain all at once.
Oh, and here’s the kicker: that’s 5× its old limit (200k tokens) and more than double GPT-5’s 400k window. And yes — it’s live right now for API users and via Amazon Bedrock and Google Cloud’s Vertex AI.
Now, here’s the interesting part:
Just last month, Anthropic was tightening the reins — slapping on new rate limits to rein in Claude Code power users.
Fast-forward to GPT-5’s launch and suddenly they’re rolling out the red carpet for devs. Because nothing lights a fire under a company like a little competition.
Why this matters:
Longer context = smarter coding. Claude can now see your entire repo, not just snippets. That means cleaner, more coherent outputs — especially for long, agentic coding sessions where it needs to remember every step
Anthropic swears it’s not just a numbers game — they’ve tuned Claude’s “effective context window” so it can actually process most of what you throw at it without getting confused. (a feat plenty of models with big context windows still struggle with)
Lastly, this keeps Claude fighting fit in the coding-model arms race, especially with OpenAI gunning for its enterprise customers.
On the competition angle:
Anthropic has built one of the biggest enterprise AI businesses, selling Claude to platforms like GitHub Copilot, Windsurf, and Cursor.
OpenAI’s GPT-5 is now the default in Cursor, and even its CEO helped announce it.
But here’s the thing: Long prompts aren’t cheap — you’re looking at $6 per million input tokens and $22.50 per million output tokens beyond the old 200k limit of $3 and $15.
But for enterprise-level teams, that’s still a solid trade-off — far better than wasting time chopping big projects into tiny chunks.
The Big picture:
This is Anthropic flexing hard in the middle of an AI coding-model turf war.
GPT-5’s pricing and performance are gunning for Claude’s throne, but this counter-move makes Claude’s coding chops a whole lot more enticing — and way harder to walk away from.
Our take: Healthy rivalry like this? It’s exactly why we get upgrades this good.
Here’s the full report.
