claude-haiku-4-5
Checking current pricing…
Chat comparison
claude-haiku-4-5vsclaude-sonnet-4-6
Compare the things that matter before you build: current price, capabilities, and latest aggregated LLM7 statistics.
Current pricing
See the cost difference clearly
claude-sonnet-4-6
Checking current pricing…
What's different
Only capabilities that differ between these models are listed. Green highlights the broader supported option or larger context window.
Context window
claude-haiku-4-5
200,000 tokens
claude-sonnet-4-6Better fit
1,000,000 tokens
Input formats
claude-haiku-4-5
text
claude-sonnet-4-6
text, image
Vision
claude-haiku-4-5
Not supported
claude-sonnet-4-6Better fit
Supported
Reasoning
claude-haiku-4-5
Not supported
claude-sonnet-4-6Better fit
Supported
Capability scores
AI-estimated ratings on 30 criteria from head-to-head comparisons against 86 other text models in the LLM7 catalog, judged by TypeSafe Jev from public information (2026-10-20). Each score is the win rate against the field, relative to the strongest model for that criterion (100%). Not a benchmark.
Below 50%50–79%80–94%95%+
claude-haiku-4-5 · leads in 1
48%
Weak
claude-sonnet-4-6 · leads in 27
84%
Strong
Knowledge & reasoning
General knowledge
claude-haiku-4-544%claude-sonnet-4-6Leads82%Logical reasoning
claude-haiku-4-523%claude-sonnet-4-6Leads84%Mathematics
claude-haiku-4-529%claude-sonnet-4-6Leads82%Scientific and technical depth
claude-haiku-4-526%claude-sonnet-4-6Leads81%Planning and task decomposition
claude-haiku-4-525%claude-sonnet-4-6Leads84%
Coding & tools
Code generation
claude-haiku-4-537%claude-sonnet-4-6Leads79%Debugging and code review
claude-haiku-4-531%claude-sonnet-4-6Leads86%Agentic coding
claude-haiku-4-532%claude-sonnet-4-6Leads89%Tool and function calling
claude-haiku-4-549%claude-sonnet-4-6Leads86%SQL and query writing
claude-haiku-4-536%claude-sonnet-4-6Leads86%
Writing & language
Creative writing
claude-haiku-4-545%claude-sonnet-4-6Leads84%Professional writing
claude-haiku-4-556%claude-sonnet-4-6Leads89%Conversational quality
claude-haiku-4-553%claude-sonnet-4-6Leads88%Multilingual ability
claude-haiku-4-544%claude-sonnet-4-6Leads73%Translation
claude-haiku-4-545%claude-sonnet-4-6Leads86%Summarization
claude-haiku-4-543%claude-sonnet-4-6Leads89%
Data & professional work
Data analysis
claude-haiku-4-533%claude-sonnet-4-6Leads87%Professional domains
claude-haiku-4-539%claude-sonnet-4-6Leads85%Information extraction
claude-haiku-4-550%claude-sonnet-4-6Leads90%Classification and labeling
claude-haiku-4-550%claude-sonnet-4-6Leads91%Long-context handling
claude-haiku-4-541%claude-sonnet-4-6Leads91%
Reliability & control
Instruction following
claude-haiku-4-555%claude-sonnet-4-6Leads88%Structured output
claude-haiku-4-570%claude-sonnet-4-6Leads93%Factual accuracy
claude-haiku-4-563%claude-sonnet-4-6Leads92%Safety and alignment
claude-haiku-4-568%claude-sonnet-4-6Leads87%Steerability
claude-haiku-4-559%claude-sonnet-4-6Leads89%Output consistency
claude-haiku-4-586%claude-sonnet-4-6Leads97%
Speed, cost & vision
Response speed
claude-haiku-4-5Leads84%claude-sonnet-4-640%Cost efficiency
claude-haiku-4-568%claude-sonnet-4-668%Image understanding
claude-haiku-4-5No image inputclaude-sonnet-4-662%
Latest LLM7 statistics
Recent aggregated percentages and response-time metrics from LLM7, not a global model benchmark or traffic disclosure.
Recent stability
claude-haiku-4-5
96.36%
claude-sonnet-4-6
97.48%
Average response time
claude-haiku-4-5
5,956.97 ms
claude-sonnet-4-6
17,871.29 ms
P95 API latency
claude-haiku-4-5
30,000 ms
claude-sonnet-4-6
60,000 ms
Side-by-side trends
Shared charts make the observed differences easier to read. Metrics without enough history stay hidden.
Request success rate
Both models on the same scale
claude-haiku-4-5: 100%claude-sonnet-4-6: 100%
Average API latency
Both models on the same scale
claude-haiku-4-5: 3,702 msclaude-sonnet-4-6: 4,447 ms
P95 API latency
Both models on the same scale
claude-haiku-4-5: 30,000 msclaude-sonnet-4-6: 5,000 ms
Estimate your workload
Published rates are loaded from the pricing API; per-request quotes are billed from actual provider usage.
Checking current pricing…
Checking current pricing…
Quick take
- claude-sonnet-4-6 has a larger context window.
- claude-sonnet-4-6 supports image input while the other model does not.
- claude-sonnet-4-6 recorded higher recent LLM7 stability during this period.
- claude-haiku-4-5 recorded a lower observed p95 API latency on LLM7 during this period.