Breaking

Anthropic Lifts Claude API Rate Limits, Ends Spend-Based Tiers

July 2, 2026 at 19:03 EDT

Foundation Models
Software Dev & Coding
AI Agents

Around June 26, 2026, Anthropic raised Claude Platform API rate limits for all users and simplified its tier structure. The latest Sonnet and Haiku models now offer up to 5x higher rate limits at the highest tier, and existing organizations were automatically moved to an equal or higher tier.

Claude Platform API · Rate Limits

Anthropic lifts API rate limits — Sonnet & Haiku now match Opus

Usage tiers are simplified from many into three — Start, Build, Scale — and at the top tier the latest Sonnet and Haiku models get 5× higher rate limits. Most organizations move up automatically, and no one gets lower limits than before.

5×

Higher rate limits for Sonnet & Haiku at the top tier

Simplified usage tiers: Start · Build · Scale

Action required — most orgs move up automatically

Prompt caching effect

An 80% cache hit stretches effective throughput 5×

nominal ITPM

10M

effective ITPM

Cache-read tokens aren't counted toward ITPM for most models (Haiku 3.5 excepted) — so caching sharply boosts real throughput.

Three tiers · monthly spend cap

Start

$500

Build

$1,000

Scale

$200,000

Limits apply per model class as RPM, ITPM and OTPM. A separate Custom tier is also available. Batches API and Managed Agents have separate limits.

Developers welcome it

More headroom for multi-agent & parallel workloads
Fewer 429 errors, greater stability in production
Simpler tiers reduce friction

But keep in mind

Nothing becomes unlimited
Real figures depend on your tier
Check your tier & limits in the Console first

Behind the move

Capacity is expanding via a partnership with SpaceX's Colossus, aligned with an Opus limit increase around May 2026 — a step in step with similar high-throughput limit raises from OpenAI and Google.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.