New Haiku 5.5: Cheaper Runs, Coding Gains, and an Effort Setting




Haiku 5.5 is on average 75% cheaper to run than Haiku 4.5, with stronger coding, computer-use, and knowledge-task performance, plus a new `effort` setting.
New Haiku 5.5: summary
- Average run cost is 75% lower than Haiku 4.5.
- Significant improvements in coding, computer use, and knowledge tasks.
- Prices are cut sharply: for tasks up to 100K tokens, the model is 90% cheaper than Haiku 4.5; for longer tasks, roughly half the price.
- The
effortsetting has been added, which earlier Haiku models did not have.
Why it matters
AnalysisLower costs and explicit effort control make Haiku 5.5 more practical for high-volume workloads and give developers finer control over price-quality tradeoffs.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Google unveils Gemini 4 Neon: model outperforms Claude Opus 5.5 and GPT-6.2 Sol on coding and agentic tasks
Google has introduced Gemini 4 Neon, a new AI model that reportedly beats Claude Opus 5.5 and GPT-6.2 Sol on most tests, especially in programming, reasoning, and agent workflows.
Thinking Machines releases Inkling 2 Mini, an open-weights model for local coding
Thinking Machines has released Inkling 2 Mini, an open-weights transformer with a cyclic architecture that can run locally on a Mac and was used to build a flight simulator offline.
Codex: Basic Onboarding, Setup, and First Tasks
There are three ways to work with AI when writing code - and Codex covers the most ambitious of them.
Token Burn Rate
A critical engineering and financial metric for the rate of consumption of contextual and generative tokens (and dollars per hour) in agent-based development sessions, factoring in prompt caching.