01Overview02Models03Rankings04Compare05Quiz06Guide

Model

Claude Haiku 4.5 NO.06

Fastest, cheapest Claude — snappy everyday tasks at scale.

CompareBack to the index

Anthropic · Claude · Fast · Closed weights

Specification

Fig.01 — Claude — Anthropic

USA

Context window
200Ktokens
Max output
64Ktokens
Input price
$1/ 1M
Output price
$5/ 1M
Cached input
$0.10/ 1M
Throughput
—
Class
Fast
Modalities
Text, Image, Pdf
Weights
Closed
Knowledge cutoff
Jul 2025
Released
Oct 2025

Overview

Claude Haiku 4.5 is Anthropic's fastest and cheapest model, built for high-volume, latency-sensitive work. It delivers surprisingly strong reasoning and coding for its price and is a great router target for the bulk of simpler traffic. Escalate to Sonnet or Opus when a task needs deeper reasoning or long-horizon reliability.

Strengths

  • Fastest and cheapest Claude
  • Strong quality for its tier
  • Great for high-volume routing
  • Low latency for real-time work
  • 200K context

Trade-offs

  • Smaller 200K context vs 1M on higher tiers
  • Not for hard long-horizon agentic work
  • Lower peak reasoning than larger Claude models

Fit

Best for

  • High-volume classification and extraction
  • Low-latency chat and autocomplete
  • Simple coding and edits
  • Cheap agent sub-steps

Not ideal for

  • Complex autonomous coding
  • Deep multi-step reasoning tasks
  • Very long-context tasks (>200K)

Capabilities

Ten axes, normalized 0–100 and scored the same way across the whole catalog — so a 78 here means what a 78 means anywhere else on the bench.

Fig.02 — Capability radar

0 — 100

Reasoning
78
Coding
78
Math
80
Writing
82
Knowledge
80
Speed
93
Agentic
75
Vision
78
Multilingual
82
Long Context
80

Benchmarks

Public results, with independent third-party runs marked. Bars normalize percentages against 100 and Elo ratings against a 1500 ceiling.

Independent index

Artificial Analysis Intelligence Index

Composite of ~9–10 independent evals · Artificial Analysis

30/ 100

BenchmarkResult
Artificial Analysis Intelligence IndexIndependentGeneral · Artificial Analysis
30%
SWE-bench VerifiedIndependentCoding · SWE-bench
73.3%
SWE-bench ProIndependentCoding · SWE-bench Pro
39.5%
AIME 2025Math · Vendor-reported
80.7%

Alternatives

Models in roughly the same class — the ones worth weighing against this record.

Compare side by side
01GPT-5.6 LunaOpenAIThe fast, cost-efficient GPT-5.6 tier for everyday work at scale.02GPT-5.4 nanoOpenAIOpenAI's cheapest small model for high-volume simple work.03Gemini 3.5 FlashGoogle DeepMindFast, cheap multimodal workhorse with near-Pro quality.

Index

PreviousClaude Sonnet 5NextGPT-5.6 Sol
All models/EDUCATION