← All models

Anthropic: Claude Sonnet 5

Anthropic

balanced

Capabilities

VisionToolsStructured OutputExtended ThinkingLong ContextCode

Strengths

  • Frontier performance across coding and reasoning tasks
  • Exceptional long-context handling with 1M token window
  • Strong multimodal vision capabilities
  • Adaptive thinking with selectable reasoning effort levels
  • Reliable structured output and tool use

Weaknesses

  • Slowest inference speed among current models (76-77 tok/s)
  • High output cost ($10/M tokens) limits batch processing
  • Limited transparency on training methodology
  • Extended thinking overhead increases latency further
  • Closed-source weights prevent local deployment

Pricing

Input / 1M tokens

$2

Output / 1M tokens

$10

Context window

1,000k

Transparency

Open weights

0.0 / 10

Open training data

0.0 / 10

Open methodology

3.5 / 10

Licence openness

1.0 / 10

Provider disclosure

4.5 / 10

FMTI company score

N/A

Composite:4.0 / 10

Closed-source model with limited public methodology disclosure. Anthropic publishes research on interpretability and constitutional AI, but training details remain proprietary. No open weights or training data.

Sustainability

Inference energy

N/A

Training footprint

N/A

Provider infrastructure

5.5 / 10

Composite:3.2 / 10

Anthropic's sustainability practices are not publicly detailed. No commitment to renewable energy infrastructure has been widely disclosed. Inference energy unknown for frontier-class model.

Task Fitness

question answering
78%
code
81%
mathematics
88%
business communication
87%
vision
83%
analysis
74%
extraction
85%
long-form generation
87%
research
87%
reasoning
68%
summarisation
92%
translation
88%
conversation
89%