Claude Haiku 5.5, launched October 7 as the third Claude 5.5 model in a month, is positioned as the low-cost, high-volume member of the family. Its headline price matches OpenAI’s GPT-6 Luna for shorter prompts, but Anthropic’s roughly 75% average cost-saving claim is more modest than the 90% cut in the listed per-token price. 1236

• Pricing: For prompts of up to 100,000 tokens, input costs $0.10 and output $0.50 per million tokens. Above that threshold, the rates are $0.50 and $2.50, respectively. The short-prompt rates are 90% below Haiku 4.5’s listed rates and match GPT-6 Luna’s; Anthropic says Haiku 5.5 costs around 75% less to run on average. Reports note that its newer tokenizer can count the same text as roughly 30% more tokens than Haiku 4.5, so a rate cut is not an identical reduction in every bill. 18125 Insufficient evidence in the retrieved material to give a reliable percentage of requests that qualify for the cheaper tier.
• Uses and positioning: Anthropic pitches it for repetitive, latency- and cost-sensitive work—including summaries, compaction, database queries, classification, extraction and routing—and for smaller assignments delegated by more capable agents using Sonnet 5.5 or Opus 5.5. It adds effort controls, though I could not verify their default setting from the retrieved evidence. 12134
• Reported performance: Anthropic reports 72.4% on the OSWorld 2.1 offline subset, versus 15.7% for Haiku 4.5, 48.9% for GPT-6 Luna and 83.9% for Sonnet 5.5. These are Anthropic-reported comparisons, not independently established rankings. I could not reliably verify the requested Terminal-Bench 4.0 figures or a like-for-like comparison with Qwen3.7 Flash and GLM-5.3-Flash; the results should not be read as proof that Haiku leads every small model. 12
• Availability and related changes: Haiku 5.5 is reported available through Anthropic’s platform, AWS, Google Cloud and Microsoft Azure. Anthropic also lists Sonnet 5.5 cached reads at $0.20 per million tokens, and says monthly API credits for Max and Team subscribers are rolling out. The retrieved evidence does not establish the prior Sonnet cached-read rate, an agentic-task savings estimate or the credit amounts, so I cannot quantify those changes confidently. 13415
• IPO context: Reuters describes the launch as preceding a planned IPO and reports substantial 2025 losses. Insufficient evidence here to state the requested marketing or listing dates, their uncertainty, or a dependable 2025 sales-and-losses pair; a planned IPO is not a confirmed listing date. 41

Claude Haiku 5.5, launched October 7 as the third Claude 5.5 model in a month, is positioned as the low-cost, high-volume member of the family. Its headline price matches OpenAI’s GPT-6 Luna for shorter prompts, but Anthropic’s roughly 75% average cost-saving claim is more modest than the 90% cut in the listed per-token price. 1236

• Pricing: For prompts of up to 100,000 tokens, input costs $0.10 and output $0.50 per million tokens. Above that threshold, the rates are $0.50 and $2.50, respectively. The short-prompt rates are 90% below Haiku 4.5’s listed rates and match GPT-6 Luna’s; Anthropic says Haiku 5.5 costs around 75% less to run on average. Reports note that its newer tokenizer can count the same text as roughly 30% more tokens than Haiku 4.5, so a rate cut is not an identical reduction in every bill. 18125 Insufficient evidence in the retrieved material to give a reliable percentage of requests that qualify for the cheaper tier.
• Uses and positioning: Anthropic pitches it for repetitive, latency- and cost-sensitive work—including summaries, compaction, database queries, classification, extraction and routing—and for smaller assignments delegated by more capable agents using Sonnet 5.5 or Opus 5.5. It adds effort controls, though I could not verify their default setting from the retrieved evidence. 12134
• Reported performance: Anthropic reports 72.4% on the OSWorld 2.1 offline subset, versus 15.7% for Haiku 4.5, 48.9% for GPT-6 Luna and 83.9% for Sonnet 5.5. These are Anthropic-reported comparisons, not independently established rankings. I could not reliably verify the requested Terminal-Bench 4.0 figures or a like-for-like comparison with Qwen3.7 Flash and GLM-5.3-Flash; the results should not be read as proof that Haiku leads every small model. 12
• Availability and related changes: Haiku 5.5 is reported available through Anthropic’s platform, AWS, Google Cloud and Microsoft Azure. Anthropic also lists Sonnet 5.5 cached reads at $0.20 per million tokens, and says monthly API credits for Max and Team subscribers are rolling out. The retrieved evidence does not establish the prior Sonnet cached-read rate, an agentic-task savings estimate or the credit amounts, so I cannot quantify those changes confidently. 13415
• IPO context: Reuters describes the launch as preceding a planned IPO and reports substantial 2025 losses. Insufficient evidence here to state the requested marketing or listing dates, their uncertainty, or a dependable 2025 sales-and-losses pair; a planned IPO is not a confirmed listing date. 41