Router DeskHomeClaude Haiku vs Sonnet
Claude Haiku vs Sonnet
Haiku is the fast, cheaper Claude. Sonnet is the one Anthropic aims at coding agents. The prices below are OpenRouter’s, checked 27 Sep 2026.
Updated 2026-09-27
| Slug | Input / M | Output / M | Context |
|---|---|---|---|
| anthropic/claude-haiku-4.5 | $1 | $5 | 200,000 |
| anthropic/claude-sonnet-4.5 | $3 | $15 | 1,000,000 |
| anthropic/claude-sonnet-5 | $2 | $10 | 1,000,000 |
Haiku 4.5’s card says it is Anthropic’s fastest Claude and that it reaches Sonnet 4-class results at lower cost and latency. Sonnet 4.5’s card aims it at agents and coding, and cites SWE-bench. Sonnet 4.5 also has a long-context override: at 200,000 prompt tokens and above, input becomes $6 and output $22.50 per million.
Which one to call
- Short classification, routing, and extraction: Haiku 4.5.
- A coding agent that stays under 200K: Sonnet 5 is cheaper than Sonnet 4.5 on both sides.
- A prompt that crosses 200K on Sonnet 4.5: budget the higher tier, or move to Sonnet 5 if your eval allows it.
- OpenRouter’s own compare URL for Haiku 4.5 and Sonnet 4.5 is the live pair.
Batch prices are half on these cards. Haiku batch is $0.50 / $2.50. Sonnet 4.5 batch is $1.50 / $7.50. Sonnet 5 batch is $1 / $5.
Questions
Is Haiku or Sonnet cheaper?
Haiku 4.5, at $1 input and $5 output per million tokens. Sonnet 4.5 is $3 and $15. Sonnet 5 is $2 and $10.
Which has the longer context?
Sonnet 4.5 and Sonnet 5 list 1,000,000 tokens. Haiku 4.5 lists 200,000.
Where is the side-by-side page?
https://openrouter.ai/compare/anthropic/claude-haiku-4.5/anthropic/claude-sonnet-4.5
Related
- Claude Sonnet 5 pricing
Claude Sonnet 5 was $2 input and $10 output per million tokens on 27 Sep 2026. Batch is $1 and $5. The slug is anthropic/claude-sonnet-5.
- Opus 4 pricing
Claude Opus 4’s card is still on OpenRouter, with no endpoint on 27 Sep 2026. Live Opus rows that day ran from $5 / $25 to $15 / $75 per million tokens.
- Claude Code on OpenRouter
Point Claude Code at https://openrouter.ai/api, put the OpenRouter key in ANTHROPIC_AUTH_TOKEN, and set ANTHROPIC_API_KEY to an empty string.
- AI model comparison
A price and context comparison of DeepSeek V4, Qwen3.7 Max, GLM 5.2, Claude Sonnet 5, GPT-5.1, and Grok 4.5 on OpenRouter, checked 27 Sep 2026.