CrossModel
Back to model catalog

Compare models

Context, limits, capabilities, price and use cases side by side — up to 4 models at a time. The URL holds your selection, so you can share it.

Claude Haiku 5.5Chat with this model
AttributeClaude Haiku 5.5anthropic/claude-haiku-5-5
Specs
Context window1M
Max output128K
Input modalitiesTextImage
Output modalitiesText
Capabilities
StreamingSupported
Tool callingSupported
Structured output (JSON)Supported
Image inputSupported
Thinking modeSupported
Price / 1M tokens
Input$ 0.10>= 100K$ 0.50
Output$ 0.50>= 100K$ 2.50
Cache read$ 0.010>= 100K$ 0.050
Cache write$ 0.13>= 100K$ 0.63
What it is good at
Best for
  • High-volume, latency-bound calls
  • Subagent for coding agents
  • Computer and browser use
Published benchmarks
FrontierCode v1.1 (Main)score
  • 46.4%effort max · Claude Code(setup not fully disclosed)
Terminal-Bench 4.0pass_rate
  • 39.2%effort max · Claude Code(setup not fully disclosed)
SWE-Bench Proresolved_rate
  • 64.8%effort max(setup not fully disclosed)
OSWorld 2.1 (offline subset, partial score)score
  • 72.4%effort max(setup not fully disclosed)
AA-Briefcase v1.1score
  • 1578effort max(setup not fully disclosed)
GDPval-AA v2.1elo
  • 1620effort max(setup not fully disclosed)
Chartography (no tools)score
  • 46.4%effort max(setup not fully disclosed)
Humanity's Last Examaccuracy
  • 45.9%effort max(setup not fully disclosed)
Humanity's Last Exam (with tools)accuracy
  • 57.4%effort max(setup not fully disclosed)