Claude Haiku 5.5: Anthropic’s New Workhorse.

Summary
The video is an overview presented by Prompt Engineering breaking down Anthropic’s release of Claude Haiku 5.5. The presenter examines the model’s pricing structure compared to Haiku 4.5 and OpenAI’s GPT-6 Luna, benchmark improvements across computer use and multidisciplinary reasoning, the new effort setting, and availability across major cloud platforms.

What is shown

• [00:00] Motion graphics announcing the release of Claude Haiku 5.5 by Anthropic.
• [00:09] Comparative price chart between Haiku 4.5 ($1.00) and Haiku 5.5 ($0.10 per 1M input tokens for prompts under 100K).
• [00:17] Side-by-side graphic comparing Haiku 5.5 ($0.10) to OpenAI GPT-6 Luna ($0.10).
• [00:24] Pricing table detailing cache reads, cache writes, input tokens, and output tokens for Haiku 5.5 versus Haiku 4.5.
• [00:30] Anthropic benchmark tables showing metrics for Knowledge work (GDPval-AA v2.1, AA Briefcase v1.1), OSWorld 2.1, and Humanity’s Last Exam.
• [00:34] Bar graph comparing computer use performance on OSWorld 2.1 across Haiku 4.5, Haiku 5.5, and Sonnet 5.5.
• [00:39] Effort level selector interface showing Low, Med (default), High, XHigh, and Max options with a cost versus intelligence trade-off chart.
• [00:48] Diagram outlining Haiku 5.5’s workhorse roles (summarization, classification, subagent tasks) alongside Sonnet and Opus, with a Terminal-Bench 4.0 comparison.
• [00:59] Timeline graphic showing the Claude 5.5 model family releases (Opus 5.5 on Sep 22, Sonnet 5.5 on Sep 28, Haiku 5.5 on Oct 07).
• [01:11] Breakdown of API credit perks for Max and Team plan subscribers.
• [01:18] Availability grid showing Claude API, AWS, Google Cloud, and Microsoft Azure (claude-haiku-5-5).

Claims & numbers

• The presenter states that Anthropic calls Haiku 5.5 the “cheapest, fastest, and most capable small model they have ever released” [00:04].
• The presenter says input pricing for most requests is $0.10 per 1M tokens (10x cheaper than Haiku 4.5’s $1.00), matching OpenAI’s GPT-6 Luna [00:09].
• The presenter states that for prompts over 100K tokens, input costs are $0.50 (half off Haiku 4.5’s $1.00) [00:20].
• The presenter notes Anthropic’s estimate that Haiku 5.5 is roughly 75% cheaper on average overall [00:26].
• The presenter reports that OSWorld 2.1 (offline subset) computer use scores jumped from 15.7% on Haiku 4.5 to 72.4% on Haiku 5.5 (compared to 83.9% for Sonnet 5.5) [00:35].
• On-screen benchmark tables state Haiku 5.5 scored 1620 on GDPval-AA v2.1 (vs. 755 for 4.5), 1178 on AA Briefcase v1.1 (vs. 814 for 4.5), 45.9% with no tools and 57.4% with tools on Humanity’s Last Exam (vs. 10.2% and 18.7% for 4.5) [00:30].
• The presenter claims Haiku 5.5 is the first Haiku model with an “effort” setting letting users adjust compute cost versus intelligence [00:39].
• The presenter mentions monthly API credit perks rolling out: $100/mo for Max 5x, $200/mo for Max 20x, and up to $500/mo pooled for Team plans [01:11].
• The presenter says Haiku 5.5 is available immediately on the Claude API, AWS, Google Cloud, and Azure [01:18].

Notable quotes

• “Anthropic just released Haiku 5.5, and they are calling it the cheapest, fastest, and most capable small model they have ever released.” [00:00]
• “It’s also the first Haiku with an effort setting, which lets you trade cost for intelligence.” [00:39]
• “Anthropic is positioning it as the workhorse model for Opus and Sonnet...” [00:45]

Assessment
This is a third-party news recap and analysis video summarizing Anthropic’s launch announcement using infographics and vendor-published benchmark tables. No live prompt execution or independent hands-on testing is conducted in the video itself.

Described by gemini-3.8-flash on 2026-10-09 from the video’s audio and frames.