Claude Haiku
by Anthropic · Model · Language & code models
mixed
Cheap and capable, but independent tests rank it behind Kimi K3 and GLM 5.3
Claude Haiku 5.5 released, big gain over previous Haiku, available on the free plan
AccessHosted app
Why we say so
Artificial Analysis ranks Haiku 5.5 behind Kimi K3 and GLM 5.3 and finds it about three times pricier than GPT-6 Luna, with MiMo 2.6 and Luna better value. Anthropic's own claim of beating Luna on agentic coding and computer use is untested by others. Rivals like Microsoft and Tencent claim to match it at lower cost, which are also maker claims. Simon Willison lists it at $0.10/$0.50 per million tokens, which sits oddly with the 'three times pricier' finding, so check current prices.
Evidence
- Oct 11Beats GPT-6 Luna on most of Anthropic's benchmarks including agentic coding and computer use
- Oct 11Ranks behind Kimi K3 and GLM 5.3 on the Artificial Analysis index; about $0.21 per task, cheaper than GLM 5.3
- Oct 11About three times pricier than GPT-6 Luna; MiMo 2.6 and Luna more efficient on intelligence versus cost