The Claude logo. Image: Anthropic
Anthropic has launched Claude Haiku 5.5, a small model it calls “the cheapest, fastest, and most capable small model we’ve ever released”, at a tenth of Haiku 4.5’s price per token. Alongside it, the company cut Sonnet 5.5’s cache prices and said Claude Max and Team subscribers will get up to $500 a month in free API credits.
Haiku 5.5 is available now in the Claude apps for Free, Pro, Max, Team and Enterprise users, in Claude Code, and to developers on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Foundry as claude-haiku-5-5.
A tenth of the price per token
For prompts up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens, against $1 and $5 for Haiku 4.5. Longer prompts cost five times as much, at $0.50 and $2.50. Anthropic says prompts under 100,000 tokens make up around 90% of requests to its previous Haiku.
Those prices match OpenAI’s GPT-6 Luna exactly. The saving is smaller than the headline numbers suggest, though: Haiku 5.5 uses Anthropic’s newer tokeniser, so the same text counts as about 30% more tokens. Anthropic says the new model “costs around 75% less to run” than Haiku 4.5 on average.
It also gets a 1 million token context window, up from 200,000, and up to 128,000 output tokens, and it is the first Haiku with an effort setting, so developers can trade intelligence against cost.
How it scores against GPT-6 Luna
In Anthropic’s own benchmarks, Haiku 5.5 beats Luna on every test both were run on:
- Computer use (OSWorld 2.1): 72.4%, against 48.9% for Luna and 15.7% for Haiku 4.5.
- Agentic coding (Terminal-Bench 4.0): 39.2%, against 16.4% for Luna. Haiku 4.5 scored zero.
- Humanity’s Last Exam: 45.9% without tools and 57.4% with them, up from 10.2% and 18.7%.
- Knowledge work (GDPval-AA): 1620, against 1437 for Luna and 1840 for Sonnet 5.5.
These are Anthropic’s figures and haven’t been independently checked. Anthropic is also clear about the limits: Sonnet 5.5 and Opus 5.5 “remain better choices for complex agentic coding tasks”, while Haiku 5.5 is meant for narrower jobs like summaries, classification, live support and subagents working for a bigger model.
Developers have to change their code
Switching isn’t a one-line change. Anthropic’s migration notes say the old manual thinking budget, custom temperature settings and prefilled assistant replies now return errors, and replies can start with thinking blocks because adaptive thinking is on by default. Requests can also come back with a “refusal” from new safety classifiers.
On safety, Anthropic says Haiku 5.5 showed “far fewer instances of misaligned behavior” than Haiku 4.5. Its cyber safeguards still block penetration testing, with vetted teams pointed to Anthropic’s Cyber Verification Program.
Free API credits for Max and Team
The bigger news for many Claude subscribers is a new monthly API credit, rolling out this week. Max 5x users get $100 a month, Max 20x users $200, and Team plans $20 per Standard seat and $100 per Premium seat, pooled and capped at $500. Free, Pro and Enterprise plans aren’t included.
Subscribers claim the credits by linking a Claude Console organisation in their billing settings, after seven days on an eligible plan. The credits cover the Claude API, Managed Agents, the Agent SDK and the playground, but not Claude Code or extra usage in the Claude apps. They expire at the end of each billing cycle and don’t roll over.
Anthropic also halved Sonnet 5.5’s cache read price to $0.10 per million tokens, which it says makes Sonnet 5.5 about 20% cheaper on most agentic work.
Why it matters
Small models do most of the grunt work inside AI products, and Haiku 5.5 lets Anthropic match OpenAI on price there for the first time in this generation while claiming a clear lead on capability. The free credits are a play for Claude’s most dedicated subscribers to start building on its platform rather than someone else’s.
Sources: Anthropic, Claude Platform docs, Claude Platform docs (API credits).
Latest Anthropic news
