OpenAI Might Have Just Killed Claude

AI Arms Race for Developers

Author

Luke Skyward
April 19, 2025

In partnership with

Find out why 1M+ professionals read Superhuman AI daily.

In 2 years you will be working for AI

Or an AI will be working for you

Here's how you can future-proof yourself:

  1. Join the Superhuman AI newsletter – read by 1M+ people at top companies

  2. Master AI tools, tutorials, and news in just 3 minutes a day

  3. Become 10X more productive using AI

Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.

OpenAI is shipping like crazy!

This week, they shipped three new models, open‑sourced a coding tool, cut prices, and started shopping for a $3 billion IDE—all in one week.

I thought Sam Altman really hates my days off.

Why the rush? Because in one corner of the AI ring, Anthropic’s Claude 3.5 / 3.7 “Sonnet” models are still every coder’s comfort blanket. Developers love Claude’s calm tone, giant context window, and its shiny new Claude Code terminal agent.

OpenAI just smashed a chair over that blanket.

Let’s see what’s under that blanket!

Claude is Developer Friendly—Here’s Why

Anthropic’s edge

  • Claude Sonnet writes neat code, follows long threads, and powers the Cursor IDE.

  • Devs say it “just listens better,” so they still pay extra.

OpenAI’s headache

  • GPT‑4o is great, but many devs swear by Claude for daily coding.

  • Losing that crowd means losing the builders who turn models into real products.

This sentiment is still visible in the webdev lmarena:

Claude still leads, but o4-mini is not yet included, and GPT-4.1 is close!

OpenAI Strikes Back: GPT‑4.1, o3, o4‑mini (launched on Apr  16 )

Model

What It Does

Why It Hurts Claude

GPT‑4.1

Stronger at coding, instruction following, and long-context tasks. Trained with better prompt steerability and excels in agentic workflows (non-reasoning).

Outperforms Claude on SWE-bench Verified (solves 55%) and offers up to 75% cheaper API pricing with caching. Ideal for agent-based dev tools.

o3

A high-performing multimodal reasoning model with support for tool use, image “thinking”, and chain-of-thought tasks.

Delivers Claude-level performance and can be used in Codex—breaking Anthropic's hold on everyday dev tools.

o4‑mini

Compact, fast, low-cost multimodal model that supports all tools (Python, search, image handling) with impressive performance-to-price ratio.

Costs less than half of Claude Sonnet’s output price, yet still competes on quality. Gives indie developers access to enterprise-grade features.

The pricing comparison - Claude’s price to intelligence edge is getting dimished with every release.

(Yes, the numbering—4.5 → 4.1—makes no sense. Even OpenAI joked about it during the launch.)

Four Big Upgrades You’ll Feel Today

  • 🛠 Tool‑First Thinking
    o‑series models can crop, rotate, and zoom images mid‑thought, call Python, then swing back to chat—all inside one answer. No plug‑ins, no copy‑paste.

Prompt: What is written on the sign?

o3 can zoom and crop the image for reasoning!

  • 💵 Brutal Price Cuts
    GPT‑4.1’s base API is up to 75 % cheaper than GPT‑4o. The mini tier dives under Claude Haiku’s price (and Haiku was supposed to be the bargain model).

  • 🖥 Codex: The Open‑Source Surprise
    An Apache‑2.0 terminal agent that edits your repo like Claude Code—but anyone can fork it (Claude Code is not open-sourced).
    Built with Ink.js; works with any model, so devs can swap Claude in—or out—with one flag.

  • 🌊 Windsurf Acquisition Talks
    OpenAI is eyeing Windsurf, an AI‑first IDE, for about $3 billion. Windsurf competes directly with Cursor. If the deal lands, the default coding home may flip from Anthropic to OpenAI overnight.

What Claude Still Has Going for It

  • Reputation for safer, calmer answers (visible chain‑of‑thought).

  • Tight AWS Bedrock integration for big‑company contracts.

  • Cursor IDE loyalty—for now.

  • But without matching price drops or opening its tools, that moat is looking shallow.

How This Affects You

If you’re a…

Why you should care

Solo hacker

o4‑mini gives near‑SOTA power for coffee‑money prices—perfect for side projects.

Team lead

Bulk caching on GPT‑4.1 can slash cloud bills. Time to re‑run your cost calculator.

Prompt tinkerer

Start writing prompts that expect web search, Python, and image steps—the models already do it.

Cursor die‑hard

Watch Windsurf. Your workflow may get a serious rival.

Heads‑up: OpenAI’s own docs admit the new models hallucinate more on some tasks—build in checks (unit tests, function‑calling, RAG, etc.).

Bottom Line: Cheaper, Smarter, and a Looming Showdown

Devs choose the tool that saves them time, money, and headaches. For a year, Claude ticked those boxes better than GPT. Today, OpenAI is back on the offensive:

If Anthropic’s only edge is developer love, we’ll take that too—by making the best dev toys and pricing them like fast food.

OpenAI Headquarters (supposedly)

Whether OpenAI kills Claude or simply scares it into faster innovation is up to Anthropic’s next move. My bet? Price cuts are coming from Claude‑land soon.

Grab your popcorn, update your API keys, and brace for lower bills.

Enjoyed this breakdown? Hit reply with your hottest Claude‑vs‑GPT take.

Thanks for reading and see you next week,

Luke