Anthropic’s October 7 announcement completes the Claude 5.5 lineup with a fast, low-cost model for high-volume work. At standard speed, Haiku 5.5 is priced at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens; Anthropic says the average cost to run it is about 75% below Haiku 4.5. The company positions it for summaries, classification, browser use, and bounded subagent tasks, while recommending Sonnet 5.5 or Opus 5.5 for complex agentic coding. (launch announcement, model documentation)

AnthropicOct 7, 2026, 18:23 UTC

The significance is economic as much as technical: a substantially more capable small model at lower cost could make routine AI processing and delegated work practical at higher volumes. Anthropic reports OSWorld 2.1 results of 72.4% for Haiku 5.5 versus 15.7% for Haiku 4.5, and Terminal-Bench 4.0 results of 39.2% versus 0%; these are company benchmark results, not evidence that Haiku replaces its larger models across tasks. (Anthropic’s benchmark comparisons)

The published price drops 90% for prompts up to 100,000 tokens and 50% above that threshold; Anthropic’s average 75% savings estimate accounts for the fact that 90% of Haiku 4.5 requests were in the cheaper tier and that the updated tokenizer uses slightly more tokens per task. Haiku 5.5 also adds adjustable effort and a 1-million-token context window, and is listed on the Claude API and major cloud platforms. (launch announcement, model documentation)

The announcement describes a Claude product release; the available evidence does not establish a connection to any $ANT token.

Ask your own

Case file

Anthropic

Answered Oct 7, 2026, 18:23 UTC

Sources 2

launch announcementanthropic.com
model documentationplatform.claude.com

Research trace 15

Latest posts of @AnthropicAI
I’m looking for Anthropic’s announcement and product documentation to establish what Haiku 5.5 changes—its price, speed, capabilities, and intended role—and to distinguish the company’s claims from demonstrated effects.
The newsroom confirms Haiku 5.5 launched on October 7 as Anthropic’s fastest, cheapest, and most capable small model, aimed at high-volume, cost-sensitive work. I’m checking the launch page and current documentation for concrete pricing, performance, availability, and how Anthropic positions it alongside Opus and Sonnet.
The previous results confirm the October 7 launch and Anthropic’s broad cost and positioning claims, but the launch page did not load. I’m checking the company’s launch post and Claude account for the specific price, capability, adjustable-effort, and availability details behind the announcement.
The launch claims point to lower cost and stronger small-model capability, but the retrieved social posts are incomplete and some benchmark figures come from users. I’m looking for Anthropic’s own model announcement and release notes to verify the feature set, prices, benchmark comparisons, and availability.
The official model overview confirms Haiku 5.5 launched October 7 with low per-token pricing, adjustable effort, a 1M-token context window, and a high-volume role. I’m checking Anthropic’s launch materials for its benchmark comparisons and the practical reason this announcement matters.
The official launch page confirms Haiku 5.5’s cheaper, high-volume role and identifies a new monthly API credit offer for Max and Team subscribers. I’m checking Anthropic’s own model card and launch benchmarks to pin down what the capability jump means and where the company still recommends using larger models.
Anthropic’s launch page and model documentation already show that Haiku 5.5 is a lower-cost model for high-volume work, with adaptive effort and a larger context window. I’m checking the official benchmark figures and whether Anthropic says it replaces larger models or complements them, to explain the practical significance accurately.
Anthropic’s official launch page and model documentation already establish the pricing, intended high-volume role, and availability. I’m checking the full benchmark table to quantify the capability change and anchor what the announcement means in Anthropic’s own comparisons.
The official launch material establishes the model’s pricing and intended role. I’m checking whether Anthropic has posted follow-up context and whether the supplied $ANT token has any documented connection to this product announcement.
The official launch details and comparison are established; this last check looks for independently reported adoption or operational updates since launch that would show whether its significance has moved beyond the product announcement.
The launch material shows a substantial price and benchmark shift, while early reports say Haiku 5.5 is already available in Claude Code and Cursor. I’m checking Anthropic’s first-party product updates to confirm rollout and clarify how the model is meant to fit into the 5.5 lineup.
Anthropic’s own page establishes Haiku 5.5 as a much cheaper option for high-volume, bounded tasks, while reserving complex agentic coding for Sonnet and Opus. I’m checking its published benchmark figures and any later rollout or pricing changes to gauge the significance beyond the headline claim.
The earlier results establish Haiku 5.5’s launch, cost, and intended role. This step checks Anthropic’s latest company posts and first-party materials for any update since launch and confirms the evidence behind its practical significance.
The official launch page and model docs establish the cost, capability, and intended high-volume role; earlier checks found no later company update or substantiated $ANT link. The key facts are now answered, so I’m closing with the product and competitive significance and its practical limits.