Claude Haiku 4.5
Multimodal
Zero-eval
#1Tau2 Retail
#1Cybersecurity CTFs
#2AIME 2025
+5 more
by Anthropic
+
+
+
+
About
Claude Haiku 4.5 continues Anthropic's tradition of fast, efficient models in the fourth generation Claude family. Designed to maintain the hallmark speed and cost-effectiveness of the Haiku line while incorporating advancements from the Claude 4 series, it serves applications requiring rapid processing and quick turnaround times.
+
+
+
+
Pricing Range
Input (per 1M)$1.00 -$1.00
Output (per 1M)$5.00 -$5.00
Providers1
+
+
+
+
Timeline
AnnouncedOct 15, 2025
ReleasedOct 15, 2025
Knowledge CutoffFeb 1, 2025
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Proprietary
Performance Overview
Performance metrics and category breakdown
Overall Performance
11 benchmarks
Average Score
69.7%
Best Score
96.3%
High Performers (80%+)
4Performance Metrics
Max Context Window
400.0KAvg Throughput
100.0 tok/sAvg Latency
0ms+
+
+
+
All Benchmark Results for Claude Haiku 4.5
Complete list of benchmark scores with detailed information
| AIME 2025 | text | 0.96 | 96.3% | Self-reported | |
| Tau2 Retail | text | 0.83 | 83.2% | Self-reported | |
| MMMLU | text | 0.83 | 83.0% | Self-reported | |
| Tau2 Telecom | text | 0.83 | 83.0% | Self-reported | |
| SWE-Bench Verified | text | 0.73 | 73.3% | Self-reported | |
| MMMU (validation) | multimodal | 0.73 | 73.2% | Self-reported | |
| GPQA | text | 0.73 | 73.0% | Self-reported | |
| Tau2 Airline | text | 0.64 | 63.6% | Self-reported | |
| OSWorld | multimodal | 0.51 | 50.7% | Self-reported | |
| Cybersecurity CTFs | text | 0.47 | 46.9% | Self-reported |
Showing 1 to 10 of 11 benchmarks
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+