01

Latest API checks

Official provider status and independent 5-minute API checks. Consumer apps may behave differently. How we measure.

Anthropic logo Anthropic / Claude

Insufficient evidence

Official API: unknown
Overall provider: unknown · fetched No verified check

Official incident coverage: unknown · Official source

Consumer components: unknown; separate from API checks.

Independent: No verified check
0 / 3 fresh regions · 0 responding

  • SeoulUnavailableLast checked No verified check · baseline insufficient
  • N. VirginiaUnavailableLast checked No verified check · baseline insufficient
  • OregonUnavailableLast checked No verified check · baseline insufficient

claude-haiku-4-5-20251001 · health-v1 · first readable text. Representative API model only; consumer apps and provider-wide health are not measured

Source: AI Fast Now independent API checks. Expected every 5 min; unavailable after a missed expected check (60 sec dispatch grace).

Pre-launch: current monitoring is not activated. No verified current snapshot is available. No visitor action starts a provider call.

Haiku is the smaller current Claude option. It uses manually enabled extended thinking, rather than the adaptive mechanism in newer Claude models.

Independent regional observations
Checking verified public snapshots. Participation alone does not establish fresh performance.
Official API model IDclaude-haiku-4-5-20251001
Release channelstable · active
Input → outputText and images → text
Context window200,000 tokens
Maximum output64,000 tokens
ThinkingManual extended thinking
Default effortEffort parameter not supported

Retirement not before 2026-10-15; no fixed retirement date announced. Official facts verified 2026-10-02.

02

Model identity and lifecycle

Family: Claude. Generation: 4.5. This entry follows the provider’s documented release channel. Official aliases: claude-haiku-4-5. Alias names are grouped into this entry to avoid duplicate model pages.

03

How to interpret performance

This model participates in the current bounded observation. Health checks and speed tests use distinct prompts and output limits. Measurements are comparable only when model revision, region, profile, prompt version, reasoning, output limit and cache policy match. Missing data never means a provider outage.

Claude consumer app performance may differ from direct API observations. No score, throughput estimate or best-time prediction is inferred from official specifications.

04

Official evidence

Anthropic model documentation · Verified 2026-10-02. Availability and lifecycle can change; the documented source governs the API contract.

All Anthropic catalog entries · Measurement methodology