Claude Opus 4.1 (August 2025)
Performance overview across all HAL benchmarks
8
Benchmarks
9
Agents
1
Pareto Optimal Benchmarks
Token Pricing
$15
Input Tokens
per 1M tokens
$75
Output Tokens
per 1M tokens
Benchmark Performance
On the Pareto Frontier? indicates whether this model achieved a Pareto-optimal trade-off between accuracy and cost on that benchmark. Models on the Pareto frontier represent the current state-of-the-art efficiency for their performance level.
| Benchmark | Agent | Accuracy | Cost | On the Pareto Frontier? |
|---|---|---|---|---|
|
Assistantbench
|
Browser-Use | 7.26% | $385.43 | No |
|
Corebench Hard
|
CORE-Agent | 51.11% | $412.42 | Yes |
|
Corebench Hard
|
HAL Generalist Agent | 35.56% | $375.11 | No |
|
Gaia
|
HF Open Deep Research | 28.48% | $1306.85 | No |
|
Scicode
|
Scicode Tool Calling Agent | 7.69% | $625.13 | No |
|
Scienceagentbench
|
SAB Self-Debug | 27.45% | $33.37 | No |
|
Swebench Verified Mini
|
SWE-Agent | 54.00% | $1789.67 | No |
|
Swebench Verified Mini
|
HAL Generalist Agent | 42.00% | $477.65 | No |
|
Taubench Airline
|
TAU-bench Few Shot | 54.00% | $294.17 | No |
|
Taubench Airline
|
HAL Generalist Agent | 54.00% | $180.49 | No |
|
Usaco
|
USACO Episodic + Semantic | 48.21% | $276.19 | No |