Breaking

Claude Haiku 5.5 Enters Code Arena WebDev at No. 30 While Matching GPT-6 Luna Pricing

October 8, 2026 at 15:53 EDT

Foundation Models
Software Dev & Coding
AI Agents

Anthropic's Claude Haiku 5.5 has entered Arena's Code Arena: WebDev leaderboard at No. 30 with a 1587 rating, putting the new low-cost Claude model nominally ahead of OpenAI's GPT-6 Luna entry while matching Luna's $0.10 input and $0.50 output price per million tokens. The result gives developers an early public signal on whether Anthropic's cheapest Claude 5.5 model can handle agentic web-development work, not just classification and summarization pipelines.

Oct. 8, 2026 · Anthropic

Haiku 5.5 lands in WebDev Arena

Anthropic's new low-cost model debuted at No. 30 with a 1587 score, matching GPT-6 Luna's $0.10/$0.50 per-million-token list price for shorter prompts.

No. 30

Code Arena rank

1587

WebDev score

$0.10

Input per 1M tokens

Context window

Input price ladder

Per 1M input tokens: Haiku 5.5 is drawn against Sonnet 5.5 and Opus 5.5 at the same scale.

$0.10

Haiku 5.5

Sonnet 5.5

Opus 5.5

The low tier is strongest below the 100,000-token prompt threshold; longer prompts move to a higher Haiku 5.5 rate.

Where it fits

Quick routing, extraction, summaries, compaction and subagent calls are the clearest production targets.

Where to be careful

Harder agentic coding still points toward Sonnet 5.5 or Opus 5.5, and high-effort runs can spend output budget on thinking.

Likely deployment pattern

Lead model plans

→

Haiku handles subtasks

→

Results return cheaply

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.