Compare Start here

AnthropicLanguage modelSuperseded

Claude Haiku 4.5

Sonnet 4-level coding at a third of the price.

Compare head-to-head Race it in 3D

Overall Intelligence

#44

of 56
21.2 pts behind #1

0

races led

0

weeks at #1

67.5

score

Released: Oct 15, 2025; 1.0 yr ago
Context: 200K; 64K max output
Price: $1 / $5; per 1M in/out · $2.00 blended
Speed: 150 tok/s; 0.8s to first token
Modalities: text · image; outputs text
Weights: Closed; API and apps

The 30-second version

Sonnet 4-level coding at a third of the price.

Key takeaways

1Sonnet 4-level coding at a third of the price.

Where it ranks

Current snapshot · illustrative scores · click any race to watch it

Haiku 4.5Best on each axis

Overall

44Overall+21.2 pts

Coding

45Coding+23.5 pts 45Frontend & UI+136 Elo 45Backend & APIs+20.7 pts 46Distributed Systems+25.1 pts 43Full-Stack Apps+20.9 pts 46Debugging & Repair+24.7 pts 43Agentic Coding+24.1 pts 45Code Review & Security+20.7 pts 45Data Science & SQL+20.3 pts 44Mobile Apps+20.6 pts 44Terminal & DevOps+27.1 pts 52Competitive Programming+789

Hard Reasoning

46Hard Reasoning+25.4 pts 50Frontier Questions+45.1 pts 45Abstract & Novel Reasoning+36.4 pts 50Logic & Puzzles+23.0 pts 45Planning & Strategy+21.8 pts

Math

53Math+27.7 pts 53Competition Math+21.6 pts 53Research Math+36.4 pts 51Applied & Quantitative+22.2 pts

Science

47Science+24.2 pts 49Graduate Science QA+15.6 pts 43Biology & Medicine+19.4 pts 48Physics & Chemistry+21.7 pts 46Research Assistance+21.7 pts

Writing & Language

47Writing & Language+23.5 pts 45Creative Writing+144 Elo 45Technical Writing+18.3 pts 48Editing & Style+18.8 pts 46Multilingual+15.3 pts 47Instruction Following+17.6 pts 46Conversation & EQ+137 Elo

Multimodal

42Multimodal+36.0 pts 38Image Understanding+21.7 pts 39Documents & Charts+20.9 pts 40Video Understanding+37.2 pts 42Speech & Audio+46.6 pts

Tool Use & Agents

41Tool Use & Agents+22.0 pts 45Tool Use+17.6 pts 32Computer Use+25.6 pts 43Web Research+28.0 pts 43Long-Horizon+26.9 hrs

Long Context

44Long Context+25.0 pts 421M Recall+20.0 pts 45Long Reasoning+23.1 pts

Speed & Cost

11Speed & Cost+28.5 pts 10Output Speed+372 tok/s 9Latency+0.45s 39Cost-Efficiency+51.4 pts

Trust & Factuality

48Trust & Factuality+24.6 pts 47Factuality+19.1 pts 47Honesty & Calibration+21.3 pts

Overall Intelligence over time

#1 frontierHaiku 4.5V4.1 FlashGLM-5.3

Update history

The distilled version first; open any release for the deeper breakdown.

The Claude Haiku family

Haiku 4.5October 2025 Haiku 5.5October 2026

60-second briefing

Haiku 4.5, explained

AI video summary · placeholder render from the auto-written script

From Anthropic's own guide, distilled

Prompt engineering and Claude best practices

Be explicit about the outcome you want; newer Claude models follow instructions literally rather than guessing at hidden intent.
Explain why a rule exists. Context about the goal lets the model generalise the rule to cases you did not list.
Wrap documents, examples and data in XML-style tags so instructions and material never blur together.
Put long documents at the top of the prompt and your question at the end; recall and grounding both improve.
For agentic tasks, state when to act versus when to ask, and ask for a short plan before large edits.

All official guides, distilled →

About Anthropic

Safety-focused lab building the Claude family of models.

Optimising for

Reliable long-horizon agentic work (especially coding and computer use), honest and calibrated answers, and strong writing, with safety research shipped alongside capability.

San Francisco, US · founded 2021

Closest rivals

Haiku 4.5 is 44th in overall intelligence; leader is Claude Opus 5.5.