Anthropic raises Claude API rate limits across all tiers, consolidates to three usage tiers
Anthropic on June 26, 2026 raised rate limits across the Claude API and collapsed its usage tier structure to three buckets — Start, Build, and Scale — eliminating the previous multi-tier ladder that set lower limits on Sonnet and Haiku than on Opus. The changes take effect immediately and require no action from existing API users.
What's new
Two changes shipped simultaneously.
Unified rate limits across model classes. Claude Sonnet 4.x and Claude Haiku 4.5 now receive the same rate limits as Claude Opus 4.x at every usage tier. Previously, Sonnet and Haiku carried lower throughput ceilings than Opus, which created mismatches for developers whose production workloads run on mid-tier models rather than the flagship.
Three-tier structure. The previous multi-tier system is now three: Start, Build, and Scale (plus a Custom tier for organizations that need more). Actual limits per tier for Opus, Sonnet 4.x, and Haiku 4.5:
| Tier | Requests/min | Input tokens/min | Output tokens/min | Spend cap/mo |
|---|---|---|---|---|
| Start | 1,000 | 2,000,000 | 400,000 | $500 |
| Build | 5,000 | 5,000,000 | 1,000,000 | $1,000 |
| Scale | 10,000 | 10,000,000 | 2,000,000 | $200,000 |
Claude Fable 5 sits on a separate, lower schedule at all tiers — 1,000 RPM / 500k ITPM on Start, 2,000 RPM / 1.5M ITPM on Build, 4,000 RPM / 4M ITPM on Scale — consistent with its position as Anthropic's most capable but most capacity-constrained model.
Tier assignments update automatically. No organization receives lower limits than before the change.
Context
Anthropic's previous tier structure had more rungs, with visible asymmetries between model classes. Sonnet 4.x — the default workhorse for most production agentic applications — ran on lower limits than Opus despite often running at higher aggregate volume across the customer base.
The simplification lands alongside two other recent API changes. On June 25, Anthropic deprecated fast mode for Claude Opus 4.7, with removal on July 24, 2026; developers using claude-opus-4-7 with speed: "fast" have until that date to migrate to Opus 4.8. And on June 15, the original Claude Sonnet 4 (claude-sonnet-4-20250514) and Claude Opus 4 (claude-opus-4-20250514) were retired entirely.
The rate limits page also carries a cache-aware ITPM note worth flagging: for most Claude models, cached input tokens do not count toward ITPM limits. With an 80 percent cache hit rate and a 2M ITPM limit, an organization can effectively process 10M total input tokens per minute. That advantage becomes more meaningful now that the same ITPM ceiling applies across Opus, Sonnet, and Haiku.
Why it matters
The rate limit equalization directly benefits teams running Sonnet 4.6 at scale — which covers most production agentic deployments. Sonnet has been the sweet spot on cost and capability for tool-use and multi-step reasoning; hitting its rate ceiling while Opus capacity remained underutilized was a friction point that sometimes pushed developers toward multi-provider setups or enterprise agreements just to reach throughput targets. That gap closes today.
The tier consolidation is a market signal. Three tiers — Start for experimentation, Build for growing applications, Scale for high-volume production — map cleanly to the lifecycle of an API-dependent product. The previous structure reflected an earlier era of cautious capacity rationing. Flattening it suggests Anthropic has enough headroom and customer clarity to operate with a simpler, more developer-friendly surface.
Corroborating sources
- Platform.claude
https://platform.claude.com/docs/en/release-notes/api
“We've raised rate limits across the Claude API. Claude Sonnet and Claude Haiku rate limits now match Claude Opus at every usage tier, and usage tiers have been consolidated into three: Start, Build, and Scale. Most organizations move to a higher tier, no organization receives lower limits than before, and no action is required.”