Claude Opus 4.6 vs GPT-5.3: The 20-Minute Model War That Crashed Software Stocks

On February 5, 2026, at approximately 9:45 AM Pacific Time, Anthropic released Claude Opus 4.6. Twenty minutes later, OpenAI dropped GPT-5.3 Codex. The near-simultaneous launch was no coincidence — both companies had originally planned 10 AM releases, but Anthropic moved its timestamp up by 15 minutes, triggering what the industry is now calling the “Model Wars of February.”

The technical achievements on both sides are genuine. Claude Opus 4.6, which Anthropic describes as a “fundamental shift in how AI handles complex workplace tasks,” achieved top scores across coding, financial analysis, and legal reasoning benchmarks. The model outperforms GPT-5.2 on a comprehensive knowledge work evaluation spanning finance, legal, engineering, and creative domains. Crucially, Opus 4.6 introduces team-of-agents architecture — coding tasks can now be distributed across multiple specialised agents working in parallel, mimicking how a human engineering team operates rather than a single developer.

OpenAI’s GPT-5.3 Codex countered with its own landmark: a 57% score on SWE-Bench Pro and 77% on Terminal-Bench 2.0, running 25% faster with fewer tokens than any prior model. Perhaps most remarkably, GPT-5.3 was “instrumental in creating itself” — OpenAI’s engineering team used early versions of the model to debug and evaluate its own performance. Alongside the model, OpenAI launched Frontier, a new platform for enterprises to build, deploy, and manage AI agents capable of executing real work.

But the real shockwave hit financial markets. Within hours of the dual launch, Anthropic’s Cowork tool released industry-specific plugins for legal, finance, sales, and marketing — and software stocks cratered. An ETF tracking the software industry suffered its worst single-day decline since April, dropping nearly 6%. Thomson Reuters fell 15.83%. LegalZoom plunged nearly 20%. The market’s verdict was blunt: if AI agents can now handle the specialised knowledge work that justified premium software subscriptions, the entire SaaS model faces existential repricing.

The competitive dynamics extend beyond models. Anthropic raised over $10 billion at a $350 billion valuation. OpenAI was valued at $500 billion late in 2025. Claude Code surpassed $1 billion in revenue just six months after its public launch. Meanwhile, OpenAI began testing advertisements inside ChatGPT on February 10, showing ads to free and Go-tier users in the US. Anthropic responded with a Super Bowl counterpunch: a 60-second pregame and 30-second in-game ad costing tens of millions, publicly committing to keeping Claude ad-free and calling advertising “incongruous” with personal AI interactions.

Prediction markets tell the rest of the story. Anthropic holds a 68% probability for best overall model by end of February. Google leads longer-term bets for March (46%) and June (54%), with Gemini 3 released in November positioning it for a potential leapfrog. The race is no longer bilateral — it’s a three-body problem with quarterly oscillations.

For India’s AI ecosystem, the model wars carry specific implications. Indian developers are the fastest-growing Claude user segment globally. Enterprise adoption is accelerating 40% quarter-over-quarter. As AI agents mature from demos to deployable systems, India’s 1.5 million annual engineering graduates represent both the largest talent pool for building on these platforms and the largest market for deploying them. The question is no longer which model wins — it’s which ecosystem India’s builders choose to build on.