Claude Haiku 5.5

Anthropic launched Claude Haiku 5.5 on October 7, 2026, positioning it as the cheapest, fastest, and most capable small model yet. The page states it costs around 75% less to run than Haiku 4.5 and is designed for summaries, compactions, database queries, and classification, while pairing well with Opus 5.5 and Sonnet 5.5 as a coding subagent. The launch page introduces a new adjustable effort setting for Haiku-class models and shows benchmark results including GDPval-AA v2.1 scores of 1620 versus 735 for Haiku 4.5, OSWorld 2.1 at 72.4%, Humanity's Last Exam at 45.9% without tools, and Terminal-Bench 4.0 at 39.2%. The Anthropic page is the creator's authoritative source for the model's capabilities.

AWS officially announced Claude Haiku 5.5 on Amazon Bedrock on October 7, 2026, marking Anthropic's newest small model as available through the Bedrock platform. The post confirms Bedrock deployment for this exact variant, giving AWS customers immediate access to the model for high-volume workloads. As a first-party AWS machine learning blog post, this serves as direct evidence that the anthropic.claude-haiku-5-5 model ID is live on Amazon Bedrock with full platform integration. Developers building on Bedrock can adopt Haiku 5.5 for cost-sensitive inference tasks using standard AWS tooling.

Anthropic's Claude Haiku family page confirms Haiku 5.5 launches on October 7, 2026, as the fastest, most cost-efficient model in the Claude 5.5 family. It is available on Claude.ai for Free, Pro, Max, Team, and Enterprise users, and on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Foundry for developers. Pricing for Haiku 5.5 on the Claude Platform is $0.10 per million input and $0.50 per million output tokens for prompts up to 100K, and $0.50/$2.50 for prompts over 100K, with up to 90% savings via prompt caching and 50% via batch. Use cases highlighted include classification, summarization, routing, compaction, real-time chat, and subagent roles.

The Claude Platform models overview lists Claude Haiku 5.5 as the fastest tier for high-volume, latency-sensitive tasks such as classification, extraction, and routing. Pricing starts from $0.10 input and $0.50 output per million tokens, with adaptive thinking, default medium effort, a 1M-token context window, and 128K max output. The docs explicitly assign the Amazon Bedrock model ID anthropic.claude-haiku-5-5 to Haiku 5.5, with a reliable knowledge cutoff of June 2026 and retirement not sooner than October 7, 2027. This is the authoritative source confirming the exact Bedrock identifier for the model in this assessment's subject.

Vals AI lists Claude Haiku 5.5 as released October 7, 2026, by Anthropic, ranking 16th of 45 models on the Vals Index at 54.31% accuracy. It posts token costs of $0.10 input and $0.50 output per million, a 1M-token context window, and 128K max output, with image and file input supported. Its best independent benchmark is 3rd of 110 on Vibe Code Bench v1.1 at 90.44%, and Terminal-Bench 4.0 scores 35.35% on Vals's own evaluation. The benchmarker's 16-model ranking, zero fallback rate, and 0.22% refusal rate give independent confirmation of the model's capabilities outside Anthropic's vendor numbers.

Beam AI's analysis details Claude Haiku 5.5 pricing as $0.10 per million input tokens and $0.50 output for prompts up to 100,000 tokens, roughly a tenth of Haiku 4.5's $1 and $5 rates. Over 100K tokens, rates jump fivefold to $0.50/$2.50, creating an easy-to-cross price step that affects subagent economics. The third-party piece reports a 1M-token context window, 128K max output, and adjustable effort, with Artificial Analysis scoring the model at 43 maximum effort and 34 at default medium effort versus 17 for Haiku 4.5. Beam's illustrative run shows 20 Haiku 5.5 subagent calls accounting for just 7% of an agent's bill, with the Opus orchestrator at 93%.