← Back to all models
C

Claude Sonnet 4.5

by Anthropic
Flagship Proprietary Multimodal 🏆 Ranked #5 of 22
86.9
Overall Score
out of 100
About

A refined iteration of Claude Sonnet 4 with improved performance on graduate-level reasoning and coding benchmarks. Claude Sonnet 4.5 delivers notably stronger results on GPQA and competitive maths whilst maintaining the same pricing as its predecessor.

Key Metrics
Context Window
200K
tokens
Avg Response
800
milliseconds
Input Cost
$3.0
per million tokens
Output Cost
$15.0
per million tokens
Arena ELO
1362
Chatbot Arena rating
MT-Bench
9.2
out of 10
Benchmark Scores
MMLU
91.7%
HumanEval
94.2%
MATH
81.5%
GPQA
67.4%
MT-Bench
92.0/10
Capability Profile
Strengths & Limitations
Strengths
✓ Best-in-class reasoning depth ✓ Excellent coding ✓ Improved GPQA over v4 ✓ Safety-focused ✓ 200K context
Limitations
⚠ Higher cost than efficient tiers ⚠ Occasional over-explanation ⚠ No image generation
Ideal Use Cases
Research assistance Advanced coding Scientific reasoning Agentic pipelines Document analysis
Model Details
Provider Anthropic
Released 2025-09-22
Open source No
Multimodal Yes
Tier Flagship
Global rank #5 / 22
Pricing (USD)
Input tokens $3.0/M
Output tokens $15.0/M
Per 1,000 tokens ≈ $0.0030 input / $0.0150 output
All Benchmarks
MMLU 91.7%
HumanEval 94.2%
MATH 81.5%
GPQA 67.4%
MT-Bench 9.2/10
Arena ELO 1362
Compare this model View leaderboard

You might also consider