Anthropic releases Claude Haiku 5.5 with 26-point intelligence gain and 75% cost cut
Anthropic released Claude Haiku 5.5, scoring 43 on the Artificial Analysis Intelligence Index, a 26-point improvement over its predecessor. The model is the first Haiku to support reasoning efforts and costs approximately 75% less to run than Claude Haiku 4.5. It features tiered pricing, matching GPT-6 Luna at $0.10/$0.50 per million tokens for prompts up to 100k tokens, with a 1-million-token context window.
IllustrationEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Reporting timeline
Anthropic releases Claude Haiku 5.5, scoring 43 on AI Intelligence Index, up 26 points from predecessor
Anthropic has released Claude Haiku 5.5, scoring 43 on the Artificial Analysis Intelligence Index, a 26-point improvement over the previous Haiku model released one year ago. Haiku 5.5 is the first Haiku model to feature Anthropic's effort settings and adaptive thinking, with tiered pricing introduced. The model costs $0.10/$0.50 per 1M input/output tokens for prompts up to 100k tokens, matching GPT-6 Luna and costing 10% of the previous Haiku model, but pricing rises 5x to $0.50/$2.50 above 100k tokens. Key performance highlights include leading small-class model performance at max effort, slightly ahead of GLM-5.3 Flash, Gemini 3.8 Flash, and GPT-6 Luna; heavy token use compared to GPT-6 Luna, using ~162k output tokens per task versus ~50k; high capability in agentic knowledge work on AA-Briefcase with 1578 Elo; and improvements on Terminal-Bench 4.0 scoring 33%, up from 0% for Haiku 4.5. The model has lower factual knowledge (36% AA-Omniscience accuracy) but a lower hallucination rate (40%) compared to peers. Its AutomationBench-AA score of 35% is considered understated due to a safety refusal issue that Anthropic is working to resolve. The context window is 1 million tokens, up from 200k for Claude 4.5 Haiku.
Read sourceAnthropic's Claude Haiku 5.5 launches on OpenRouter with reasoning support, 75% cost cut
OpenRouter announced that Anthropic's Claude Haiku 5.5 is now live on its platform. This is the first Haiku model to support reasoning efforts. According to OpenRouter, the new model is approximately 75% cheaper than its predecessor, faster with measured speeds exceeding 100 tokens per second, and shows a clear capability boost in benchmarks. The announcement was made via a post on X (formerly Twitter) with a link to use the model.
Read sourceAnthropic releases Claude Haiku 5.5, claiming 75% lower cost than previous version
Anthropic announced the release of Claude Haiku 5.5, describing it as the cheapest, fastest, and most capable small model the company has ever released. According to the announcement, the new model costs on average around 75% less to run than its predecessor, Claude Haiku 4.5. The post includes a link to further details. The announcement positions the model as a cost-efficient option for users seeking high performance from a smaller AI model.
Read source