By the toolist.ai editors. Researched from public sources, not hands-on tested by toolist.ai. Updated October 13, 2026; prices checked October 13, 2026.

Claude Haiku 5.5 vs Claude Sonnet 5.5

Verdict

Use Claude Sonnet 5.5 for coding and agent work, and Claude Haiku 5.5 for high-volume, narrowly scoped jobs. Haiku costs a twentieth of Sonnet's token price on prompts up to 100,000 tokens ($0.10 and $0.50 against $2 and $10 per million), but Sonnet scores higher on all eight Artificial Analysis benchmarks both have, including 68.4 against 36.5 on the Coding Agent Index. Above 100,000 prompt tokens Haiku's price rises fivefold, so Sonnet costs four times as much rather than twenty.

Verdict based on published prices, limits and benchmarks, not on our own tests.

Choose Claude Haiku 5.5 if

  • You run high-volume, latency-sensitive jobs such as classification, extraction or routing
  • Your prompts stay under 100,000 tokens: the support chatbot workload costs $50 a month on Haiku against $1,000 on Sonnet
  • You want a fast subagent for summaries or compaction under Sonnet or Opus, which Anthropic suggests
  • Speed matters more than depth (Artificial Analysis measures 168.0 output tokens a second at medium effort against 96.8 for Sonnet at high)

Choose Claude Sonnet 5.5 if

  • You do everyday coding and agent work, where Sonnet scores 68.4 against 36.5 on the Coding Agent Index
  • Your prompts often pass 100,000 tokens, since Sonnet has no long-context surcharge
  • You need the stronger reasoning model at the everyday setting (46.8 against 34.5 on the Intelligence Index)

Price

Input costs 20 times as much on Claude Sonnet 5.5 as on Claude Haiku 5.5. Output costs 20 times as much on Claude Sonnet 5.5 as on Claude Haiku 5.5.

Per million tokensClaude Haiku 5.5Claude Sonnet 5.5
Input$0.10 source$2 source
Cached input$0.01 source$0.10 source
Output$0.50 source$10 source

Claude Haiku 5.5: Prompts over 100,000 tokens are billed at $0.50 in, $2.50 out and $0.05 for cache reads per million tokens for the whole request; the prompt length counts cache reads and writes. Cache writes cost $0.125 per million tokens (5-minute cache) or $0.20 (1-hour cache), or $0.625 and $1 for prompts over 100,000 tokens. The Batch API costs half.

Claude Sonnet 5.5: Cache writes cost $2.50 per million tokens (5-minute cache) or $4 (1-hour cache). The Batch API costs half. There is no long-context surcharge and no fast mode.

Worked example: what a month costs

We priced three monthly workloads at the standard rates. The provider does not publish these figures.

WorkloadClaude Haiku 5.5Claude Sonnet 5.5
Support chatbot
100,000 replies, each 3,000 input and 400 output tokens, no caching (300M in, 40M out)		 $50.00$1,000
Document Q&A
5,000 questions over the same 60,000-token document set (90% cache hits), 800 output tokens each (300M in, 4M out)		 $7.70$127
Coding agent
200 tasks, each 1.5 million input tokens (90% cache hits) and 50,000 output tokens (300M in, 10M out)		 $10.70$187

Price your own workload

ModelCost a month
Claude Haiku 5.5 $7.50
Claude Sonnet 5.5 $150

Standard API prices, no batch or priority discounts. Cached input is billed at the cache-read price; cache-write charges are not included. Reasoning models bill their thinking as output tokens.

Limits and release dates

Claude Haiku 5.5Claude Sonnet 5.5
ReleasedOctober 7, 2026 sourceSeptember 28, 2026 source
Context window1,000,000 tokens source1,000,000 tokens source
Maximum output128,000 tokens source128,000 tokens source
Knowledge cutoffJune 2026 sourceJune 2026 source
Input and outputText and images in; text out sourceText and images in; text out source
API model nameclaude-haiku-5-5claude-sonnet-5-5

Benchmarks

Scores as published by the provider or measured by Artificial Analysis; higher is better. A dash means no score was published.

BenchmarkClaude Haiku 5.5Claude Sonnet 5.5
Artificial Analysis Intelligence Index
max effort		 43.4 source56.0 source
Intelligence Index at everyday effort 34.5 (medium effort, the API default) source46.8 (high effort, the API default) source
Coding Agent Index
Claude Code harness, max effort		 36.5 source68.4 source
Terminal-Bench 4.0
max effort		 32.8% source63.6% source
Humanity's Last Exam
max effort		 44.4% source55.0% source
GDPval-AA v2.1 (Elo)
max effort		 1620 source1838 source
SciCode
max effort		 55.0% source61.0% source
Terminal-Bench-Science 0.1
max effort		 20.0% source53.3% source
Output speed 168 tokens/s (medium effort, Anthropic API) source96.8 tokens/s (high effort, Anthropic API) source

Benchmark and speed data from Artificial Analysis where linked.

Read more

Sources