Anthropic launched Claude Haiku 5.5 on Wednesday, October 7, 2026, calling it the cheapest, fastest, and most capable small model it has released. The official announcement centers the model on high-volume, low-latency work such as classification, extraction, routing, and subagent tasks.
On the Claude Platform, the model ID is claude-haiku-5-5. The context window reaches 1 million tokens and maximum output is 128,000 tokens. For prompts up to 100,000 tokens, input costs $0.10 per million tokens and output costs $0.50. Above that threshold, the rates rise to $0.50 and $2.50. Anthropic says the model costs about 75% less to run on average than Haiku 4.5, and that about 90% of requests to the previous Haiku stayed under 100,000 tokens.
What changes in practice
Haiku 5.5 is the first model in the line with an adjustable effort parameter and adaptive thinking. The company positions it alongside Opus 5.5 and Sonnet 5.5 in coding workflows, especially as a subagent, and for speed-sensitive uses such as live support and browser use.
There is a cost detail that does not show up in the price table alone. The model uses the same newer tokenizer as Claude models from the 4.7 generation onward. Anthropic’s documentation says the same text can count as about 30% more tokens than on Haiku 4.5. Anyone migrating needs to look at real volume, not only the price per million tokens.
In the same release, Anthropic halved the price of Sonnet 5.5 cache reads. The company says that makes Sonnet 5.5 about 20% cheaper on most agentic work.
Where it is available
Free, Pro, Max, Team, and Enterprise users can select Haiku 5.5 on Claude.ai, on the web and on iOS and Android. For developers, the model is on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Foundry, as well as in Claude Code. Claude’s official X account also cited availability on Microsoft Azure.
In benchmarks published by Anthropic itself, Haiku 5.5 scores 1620 on GDPval-AA v2.1, against 735 for Haiku 4.5 and 1437 for GPT-6 Luna, which the company uses as a reference. On computer use, in the offline subset of OSWorld 2.1, the reported score is 72.4%, compared with 15.7% for Haiku 4.5 and 48.9% for Luna. These are the maker’s figures, not an independent evaluation.
The official account also said Haiku 5.5 shows broad improvement on internal alignment evaluations relative to Haiku 4.5, with fewer cases of misaligned behavior. That comparison is also Anthropic’s.
Sources
Transparency: This content was created, edited, or reviewed with the assistance of artificial intelligence. The information was cross-checked with public posts on X and sources available on the internet. Check the original sources for the full context.
Por GeekikiBot