Claude Opus 4 and Sonnet 4 launched May 22, 2025 at $15/$75 and $3/$15 per million tokens, hitting 72.5% and 72.7% on SWE-bench Verified. That's the short answer. The longer answer is how that one release kicked off the fastest model-war cadence in AI history and helped carry Anthropic to a $965B valuation thirteen months later.
The anthropic claude 4 release matters less for what it could do in isolation and more for what it started. Every subsequent Claude and GPT release since has been measured against the SWE-bench and pricing bar that Opus 4 and Sonnet 4 set. I pulled the original launch numbers, the pricing since, and where the frontier actually sits now that we're several generations past Claude 4.
Sources: Anthropic's official Claude 4 launch announcement (May 2025), OpenRouter model pricing pages, Anthropic Series G/H funding announcements, checked July 2026.
What the Anthropic Claude 4 Release Actually Shipped
The anthropic claude 4 release, on May 22, 2025, introduced two models — Claude Opus 4 and Claude Sonnet 4 — built specifically around long-horizon coding and agentic tool use rather than general chat quality. Opus 4 scored 72.5% on SWE-bench Verified and Sonnet 4 scored 72.7%, a then-state-of-the-art result that beat every prior Claude generation and most competing models available at the time.
Pricing was the other half of the story: Opus 4 launched at $15 per million input tokens and $75 per million output tokens, while Sonnet 4 came in far cheaper at $3 and $15. That price gap established Anthropic's now-familiar two-tier structure — a premium reasoning model for the hardest agentic tasks, and a cheaper, near-equally-capable model for everyday production workloads.
Claude 4 vs. GPT-5: How the Comparison Actually Played Out
Claude 4 actually shipped before GPT-5 — Anthropic's May 2025 release predates OpenAI's GPT-5 launch by several months, so the earliest head-to-head comparisons pitted Opus 4 and Sonnet 4 against GPT-4o and o3, not GPT-5. Once GPT-5 arrived, the rivalry became a genuine back-and-forth rather than a one-sided win for either lab.
| Release Era | Anthropic Model | OpenAI Counterpart | Reported Edge | Approx. Pricing (in/out per 1M) |
|---|---|---|---|---|
| May 2025 | Claude Opus 4 / Sonnet 4 | GPT-4o / o3 | Claude led SWE-bench Verified (72.5-72.7%) | $15/$75 (Opus 4) |
| Late 2025 | Claude 4.5 series | GPT-5 (initial) | Split — GPT-5 stronger on general knowledge and math | $15/$75 (Opus 4.5) |
| Feb 2026 | Claude Opus 4.6 | GPT-5.4 | GPT-5.4 led BenchLM's blended leaderboard, 94 to 92 | $15/$75 (Opus 4.6) |
| Apr 16, 2026 | Claude Opus 4.7 | GPT-5.4 | Anthropic reclaimed coding lead — 64.3% SWE-bench Pro | ~$15/$75 |
| Apr 23, 2026 | Claude Opus 4.7 | GPT-5.5 | OpenAI answered five weeks later with GPT-5.5 | $5/$30 (GPT-5.5) |
| May 28, 2026 | Claude Opus 4.8 | GPT-5.5 | Anthropic fired back again within five weeks | $5/$25 (Opus 4.8) |
| June 2026 | Claude Sonnet 5 | GPT-5.5 | Anthropic shifted flagship cost-performance to Sonnet tier | Below Opus 4.8 pricing |
Figures are 2025-2026 releases blended from Anthropic and OpenAI launch announcements, BenchLM.ai, and OpenRouter pricing pages. SWE-bench Pro and SWE-bench Verified are related but distinct benchmark suites; scores are not directly interchangeable across rows.
Claude Opus 4.8 vs. GPT-5.5: API Pricing per Million Tokens (Mid-2026)
OpenRouter and CloudZero pricing pages, checked June 2026
Where Anthropic's Claude 4 Bet Led by Mid-2026
The coding-agent focus Anthropic bet on with the Claude 4 release paid off commercially. Anthropic closed a $30B Series G at a $380B post-money valuation on February 12, 2026, then a $65B Series H at $965B post-money by May 2026 — a 2.5x valuation jump in under three months. Run-rate revenue climbed from roughly $9B at the end of 2025 to an estimated $45-47B by May 2026, which Anthropic's own leadership described as 80x annualized growth, adding roughly $96M in new annualized revenue per day at peak velocity.
Anthropic has also filed confidentially for an IPO as of June 1, 2026, at that $965B valuation. None of that traces to Claude 4 specifically two generations later, but the coding and agentic-use benchmarks Opus 4 and Sonnet 4 set in May 2025 are the foundation every subsequent release — 4.5, 4.6, 4.7, 4.8, and now the Sonnet 5 family — has had to beat. If you're tracking how these valuations compare across the AI stack, our AI Valuations dashboard tracks funding rounds and multiples across labs, and Big Tech Earnings tracks how Microsoft and Google are pricing their own model bets against Anthropic's.
Bottom line: the anthropic claude 4 release on May 22, 2025 — 72.5% and 72.7% SWE-bench scores at $15/$75 and $3/$15 pricing — is no longer Anthropic's current model, having been superseded by four-plus major releases through Claude Opus 4.8 and Claude Sonnet 5 by mid-2026. But it's the release that proved coding-and-agent focus could out-earn general-chat breadth, a bet that helped take Anthropic from roughly $9B to $45-47B in run-rate revenue and to a $965B valuation in about seventeen months. Explore how that plays out across the broader market on Value Add VC.
Get VC data most people never see — free.
Weekly benchmarks, valuations, and fund data. No spam, unsubscribe anytime.