claude-haiku-4-5

Anthropic announced Claude Haiku 4.5 on October 15, 2025, positioning it as the company's latest small model available to all users, with the company explicitly stating that "what was recently at the frontier is now cheaper and faster." The launch announcement directly names the exact variant and frames it as delivering coding performance comparable to Claude Sonnet 4 at one-third the cost and more than twice the speed, while also surpassing Sonnet 4 on certain tasks such as computer use. Claude Haiku 4.5 is priced at $1 per million input tokens and $5 per million output tokens, and is available via the Claude API for developers. The launch post emphasizes multi-agent orchestration use cases, noting that Claude Sonnet 4.5 can decompose complex problems into multi-step plans and orchestrate multiple Haiku 4.5 instances to execute subtasks in parallel, and that the model is well suited to real-time, low-latency applications such as chat assistants, customer service agents, and pair programming. Third-party quotes from Warp and Augment embedded in the page describe Haiku 4.5 as a leap for agentic coding, with Augment's agentic coding evaluation reportedly placing it at 90% of Sonnet 4.5's performance. The model is positioned by Anthropic as a sub-agent and computer-use specialist complementing the larger Sonnet 4.5, not a replacement for it.

The hidekazu-konishi.com article, first published May 16, 2026 and last updated September 3, 2026, is a third-party personal blog compiling a complete Anthropic Claude model release timeline from Claude 1 in March 2023 through Claude Fable 5 and Claude Sonnet 5 in June 2026 and Claude Opus 5 in July 2026. It summarizes This timeline is a curated single-author reference rather than an official Anthropic source, and it offers family-level historical context rather than novel Haiku 4.5 news, with no 302.AI-specific content. Its value lies in situating Haiku 4.5 within the broader Claude release chronology and confirming cross-platform a

The benchmarks.company model card for Claude Haiku 4.5 explicitly names the exact variant as an Anthropic-developed proprietary model with a 200K context window, released October 2025, and replicated Anthropic's published pricing of $1 per million input tokens and $5 per million output tokens. The page serves as a sourced, dated reference aggregating benchmark scores with provenance labels distinguishing developer-reported from third-party measurements. The aggregator lists SWE-bench Verified at 73.3% resolved (developer-reported, 2025-10), Terminal-Bench 2.0 at 28.3% tasks resolved (third-party, 2025-10), LiveCodeBench at 62% pass (developer-reported, 2026-06), tau-bench at 83.2% pass (developer-reported, 2025-10), GPQA Diamond at 73% accuracy (third-party, 2025-10), AIME 2025 at 80.7% solved (developer-reported, 2025-10), and MMLU-Pro at 80% accuracy (developer-reported, 2026-06). The page is a standing reference rather than breaking news, but the supplied excerpt confirms the exact variant name, context length, pricing, release date, and each cited score with its stated provenance.

BenchLM's aggregator profile for Claude Haiku 4.5, dated September 4, 2026, records the model as released October 15, 2025 with a 200,000-token context window, API pricing of $1 input and $5 output per million tokens, and cached input at $0.10 (with batch-plus-cache pricing noted at $0.050 and blended pricing at $3). T BenchLM provides up-to-date comparative context as of September 2026 but is a third-party aggregator with no announcement content and no 302.AI-specific evidence. Its pricing and context-window figures corroborate CNBC and Caylent, while the composite capability ranking adds a current-catalog perspective absent from th

n8n's AI benchmark profile for Claude Haiku 4.5 describes it as Anthropic's fastest and most efficient model, delivering near-frontier intelligence at lower cost and latency than larger Claude models. It confirms a 200,000-token context window and reports a 73% score on SWE-bench Verified, positioning Haiku 4.5 among t The n8n page is a third-party benchmark aggregator rather than an official Anthropic source, and it contains no announcement or release news and no 302.AI-specific material. Its value lies in independent benchmark scoring and pricing confirmation that aligns with other sources (200K context, $1/$5 pricing, ~73% SWE-ben