Anthropic launches Claude Haiku 5.5: 90% price cut matches GPT-6 Luna rates

#What happened

Anthropic launched Claude Haiku 5.5 on October 7, 2026, the third model in the Claude 5.5 family after Opus 5.5 (September 22) and Sonnet 5.5 (September 28). Built for high-volume, cost-sensitive work like classification, summarization, and extraction, plus live support, voice agents, and in-app assistants.

#The pricing

$0.10 per million input tokens and $0.50 per million output tokens for requests under 100,000 tokens — a 90% cut from Haiku 4.5 ($1.00/$5.00). Requests above 100,000 tokens cost $0.50 input / $2.50 output (a 50% cut). The lower tier matches OpenAI's GPT-6 Luna exactly. Anthropic estimates roughly 75% average workload savings versus Haiku 4.5, after accounting for an updated tokenizer that uses somewhat more tokens for equivalent work.

#Benchmarks

Anthropic's vendor-reported results (not independently verified): Terminal-Bench 4.0 agentic coding 39.2% versus 16.4% for GPT-6 Luna and 0.0% for Haiku 4.5, scored at maximum effort (the default medium effort scores around 20%); OSWorld 2.1 offline computer operation 72.4% versus 48.9% for Luna; GDPval-AA v2.1 knowledge work 1,620 versus 1,437 for Luna. Haiku 5.5 introduces adjustable effort levels, with medium as the default.

#Availability

API identifier claude-haiku-5-5, launching through Anthropic's own platform plus AWS, Google Cloud, and Microsoft Azure. Python and TypeScript SDK updates add beta capabilities for operating browsers and computers. Per Reuters, it is Anthropic's first Haiku with built-in safeguards for a narrow set of high-risk cybersecurity requests, with most everyday tasks unaffected.

#Also in the launch

Sonnet 5.5 cache reads drop from $0.20 to $0.10 per million tokens, an estimated ~20% saving on typical agent workloads. Monthly API credits roll out this week: $100 for Max 5x subscribers, $200 for Max 20x, and up to $500 shared among a Team subscription's users.

#Why it matters

The small-model price war is entering a new round: GPT-6 Luna, Gemini 3.5 Flash-Lite, and Grok 4.3/4.7 now face a Haiku at the same $0.10/$0.50 tier for short requests. Anthropic positions Haiku as the supporting agent worker in enterprise stacks, running the cheap repetitive turns while Opus and Sonnet handle the hard work.

#Sources

• Reuters: https://www.reuters.com/business/anthropic-launches-third-claude-55-model-expanding-ai-lineup-before-planned-ipo-2026-10-07/
• VentureBeat: https://venturebeat.com/technology/anthropic-launches-claude-haiku-5-5-with-90-api-price-reduction-matching-gpt-6-luna

Verified 2026-10-07.