N
NexusRoute
Back to Models

Claude Haiku 4.5

Anthropicbudget

Anthropic's fast and affordable model with industry-leading safety for its tier. Matches the original Claude 3.5 Sonnet on many tasks at a fraction of the cost. The go-to Anthropic model for high-volume safety-conscious deployments.

Released 2025-10-15Knowledge cutoff: 2025-07
Medium confidence|Updated 87d ago|88% source confidence

Specifications

Context Window

200K tokens

Max Output

16.4K tokens

Input Price

$0.800 / 1M tokens

Output Price

$4.00 / 1M tokens

Latency Tier

Ultra Fast (speed score: 9/10)

Capability Profile

Safety & Enterprise
9.5/10
Speed
9/10
Structured Output
8.5/10
Instruction Following
8.5/10
Cost Efficiency
8/10
Tool Use
8/10
Coding
7.5/10
Long Context
7.5/10
Factuality
7.5/10
Conversational
7.5/10
Reasoning
7/10
Creativity
6.5/10
Multimodal
6/10

Feature Support

Vision Yes
Audio In No
Audio Out No
Video No
Image Generation No
Image Editing No
Function Calling Yes
JSON Mode Yes
Structured Output Yes
Streaming Yes
Reasoning No
Realtime No
Computer Use No
Web Search No

Best Use Cases

Safety-critical enterprise chatbots and customer-facing applications
High-volume classification, extraction, and routing tasks
Document summarization at scale with 200K context
Content moderation and policy-adherent text generation
Quick code explanation and lightweight code generation

Not Ideal For

Complex software engineering (Sonnet/Opus are significantly better)
Deep multi-step reasoning or mathematical proofs
Creative writing requiring depth and nuance
Audio or video processing

Strengths

Best safety alignment of any budget-tier model — safe enough for regulated industries
Fast inference with consistent quality
200K context at budget pricing with good recall
Reliable structured output and function calling
Matches Claude 3.5 Sonnet (the 2024 leader) on many routine tasks

Weaknesses

Large quality gap vs Sonnet 4.6 on complex coding and reasoning
More expensive than GPT-5.4-nano for similar-quality simple tasks
No audio or video processing
Can be overly cautious — inherits Anthropic's conservative safety profile
Creative output lacks personality compared to larger Claude models

Edge Cases & Notes

May refuse requests that GPT-5.4-nano or Gemini Flash would handle, due to stricter safety
Prompt caching makes it very cost-effective for repetitive system-prompt workflows
Quality on code is good but not exceptional — fine for code explanation, weak for complex generation

Provider Notes

Ideal when Anthropic's safety properties are a hard requirement at lower cost. Available through Anthropic API, Amazon Bedrock, Google Vertex AI. Prompt caching recommended.

Benchmarks

MMLU84.2%
HumanEval86%
Arena Elo1280

Benchmark Notes

MMLU-Pro 84.2%. Roughly matches the original Claude 3.5 Sonnet benchmarks. Strong for a budget model, especially on safety evaluations where it leads its tier.

Research Meta

Last Evaluated

2026-03-01

Source Confidence

88%

Evaluation Method

LMSYS Arena, MMLU-Pro, safety evaluations, cost-quality analysis

Needs Re-evaluation

No

Sources

  • Anthropic Claude Haiku 4.5 model card
  • LMSYS Chatbot Arena
  • Anthropic safety evaluations

Continue exploring

Route a prompt

See how Claude Haiku 4.5 ranks

Compare models

Side-by-side analysis

Browse registry

Explore all 24 models