SHELFCRITTER AI Digest

MODEL

Claude Haiku 5.5

modeltopic-noteanthropic

Overview

Claude Haiku 5.5 is Anthropic‘s small-tier release in the Claude 5.5 generation, shipped on 2026-10-07 hours before OpenAI‘s GPT-6 launch day. It is the first Haiku-line model to expose an effort setting, bringing routing-tier parity with the Opus and Sonnet classes. Pricing is tiered on prompt length — $0.10/M input and $0.50/M output at ≤100K tokens, $0.50/M input and $2.50/M output at >100K tokens — a 10× cut on short-prompt input versus the newly-legacy Claude Haiku 4.5 at $1/$5. The ≤100K input tier now matches The Decoder’s reported GPT-6 Luna / Decisions API rate, collapsing the small-tier input-price floor across the two frontier US labs.

Timeline

  • 2026-10-08-AI-Digest — Anthropic ships Claude Haiku 5.5 on 2026-10-07, hours before OpenAI’s GPT-6 drop. First Haiku to expose an effort setting; pricing tiered on prompt length ($0.10/$0.50 at ≤100K tokens, $0.50/$2.50 at >100K). Claude Haiku 4.5 is explicitly reclassified as “legacy” at $1/$5 — the short-prompt input tier has fallen 10× for the current Haiku generation, and the $0.10/M floor now matches the reported GPT-6 Luna / Decisions API input rate. Load-bearing softener: the Oct 7 landing hours before GPT-6 is consistent with counter-programming, but two labs releasing in one day is roughly +1σ versus the ~1-frontier-item/day baseline, not clear evidence of coordinated clustering.
  • 2026-10-09-AI-Digest — Haiku 5.5 reappears today as the price-performance substrate under Anthropic‘s new Claude Dashboards / Claude Motion workload surfaces — “adjustable effort on a $0.10/M input tier (≤100K prompts) is the price-performance substrate these workload surfaces run on.” Separately, HN commenters push back on the “DeepSeek V4.1 Flash structurally undercuts frontier labs on cost” frame by citing Haiku 5.5’s pricing reset earlier this week as a direct counter-datapoint.

Key Developments

  1. First Haiku Effort Setting (October 7, 2026): Haiku 5.5 is the first in the line to expose an effort setting, bringing routing-tier parity with the Opus and Sonnet classes. The practical consequence is that effort controls, not headline input price, are the real competitive edge at the small tier.

  2. Prompt-Length-Tiered Pricing With 10× Short-Prompt Input Cut vs Haiku 4.5 (October 7, 2026): Pricing on the primary Anthropic page is $0.10/M input and $0.50/M output at ≤100K tokens, $0.50/M input and $2.50/M output at >100K tokens. Claude Haiku 4.5 is now explicitly reclassified as “legacy” at $1/$5 — the short-prompt input tier has fallen 10× for the current Haiku generation.

  3. Small-Tier Input-Price Floor Collapses to $0.10/M Across Frontier US Labs (October 7, 2026): The Haiku 5.5 ≤100K input tier matches the price The Decoder reports for OpenAI‘s GPT-6 Luna / Decisions API. Headline input-cost differentiation at the small tier is now effectively zero, with real routing-tier decisions moving to output rate (Haiku’s $0.50/M vs Luna’s reported free-output), effort controls, and provider-specific primitives.

See also: Anthropic, Claude Haiku 4.5, Claude, GPT-6 Luna, MOC - Major Companies, MOC - Developer Tools.