Anthropic's $965B Valuation Lands the Same Week Its Cheapest Model Gets Outpriced

Anthropic is racing toward a trillion-dollar IPO while Haiku now costs more than GPT-5.4 Mini. The vendor pecking order is being rewritten in real time, and pricing is the tell.

Anthropic just raised $65 billion at a $965 billion post-money valuation, per TechCrunch, in what's likely its last private round before going public. The same week, developer Simon Willison pointed out on X that Anthropic's small-tier Haiku now costs more than GPT-5.4 Mini and several Gemini models. Read together, those two data points sketch the awkward middle the company is sitting in — premium positioning at the top, eroding price competitiveness at the bottom.

The pricing squeeze is real

Willison's note has teeth because it's specific. Opus dropped from $15/$75 per million tokens in the Opus 4.1 era to $5/$25 from Opus 4.5 onward in November 2025, but Haiku — the model developers actually batch — has not kept pace with the rest of the market. Meanwhile Mistral shipped Mistral Small 4 on March 16, 2026, unifying reasoning, multimodal vision, and agentic coding into one model, and Microsoft added Mistral Medium 3.5 to Copilot Studio on May 28, 2026 with in-region data control. The cheap, fast tier is getting crowded with options that didn't exist a year ago.

OpenAI keeps shipping surface area

OpenAI's move this week was distribution, not pricing. Codex computer use now works on Windows, with task control from the ChatGPT mobile app — a meaningful expansion for the enterprise developer base still running Windows fleets. Separately, OpenAI announced Rosalind Biodefense, expanding GPT-Rosalind access for U.S. government and allied partners. The pattern: OpenAI is colonizing operating systems and federal procurement channels while Anthropic raises money.

Implications

For builders, the practical takeaway is that the small-model tier is now a real shopping decision — Haiku is no longer the default by virtue of being cheap. Benchmark your batch workloads against Mistral Small 4, GPT-5.4 Mini, and Gemini's small-tier offerings before renewing commitments. For buyers evaluating Anthropic post-IPO, the $965B valuation prices in continued frontier dominance, not unit economics at the volume tier. For creators and PMs picking a default chat partner, Anthropic still leads on quality at the top end — Opus 4.5's pricing cut suggests they know where their moat actually is.

The bottom line: Rerun your model pricing math this week — Haiku's relative position has shifted, and the small-tier choice is no longer reflexive.

Sources