Claude Haiku 5.5 Launches at $0.10 per Million Input Tokens, but Price Jumps 5x Past 100K: Do the Math Before You Migrate

Haiku 5.5's $0.10 only covers prompts up to 100K, and with a tokenizer that counts about 30% more, your old 77K prompt is already standing at the price cliff.

Marcus Ide · October 10, 2026

Anthropic launched Claude Haiku 5.5 (model ID claude-haiku-5-5) on October 7, 2026, describing it as its cheapest and fastest small model to date, aimed at summaries, context Compaction in Claude Code, classification and Subagent tasks. Claude Code v2.1.293 made it the default Haiku model on the Anthropic API the same week.Specs and Pricing: 100K Is the Dividing LineHaiku 5.5 offers a 1M-Token Context Window and up to 128K output tokens (batch jobs reach 300K in beta), takes text and images...

practice

Claude Code Hooks Add onFailure: "block": A Mistyped Script Path No Longer Lets Your Policy Gate Silently Pass Everything

When a policy hook breaks, the default is to let everything through, so your...

practice

Claude Code's Agent Tool Gets an effort Parameter: One Subagent Type, No More Separate Copies for Every Thinking Depth

Keeping one copy of a subagent per effort level was a workaround for the...

practice

Claude Code Adds --max-findings: Stop Code Review From Dumping a Hundred Suggestions on You at Once

A code review that dumps a hundred suggestions at once isn't thorough, it's...

Useful Resources

Claude API Status → Model Pricing → Prompt Playground → Token Counter → MCP Servers → LLM Benchmarks → Model Comparison →

Glossary

View All →

Agent Permission Scoping

The practice of deliberately granting an <a href="/en/glossary/core-concepts/ai-agent/">AI Agent</a> only the minimum set of permissions needed for its current task, rather than granting broad access far beyond actual need for convenience, so the potential impact of an error stays as contained as possible.

What is <a href="/en/glossary/agent-permissions/agent-permission-scoping/">Agent Permission Scoping</a>, and how does it differ from the intuition...

advanced

Least-Privilege Agent

Granting an agent only the minimum set of tools and data access strictly needed for its current task, rather than starting broad and narrowing later.

What does the principle of least privilege actually mean when applied to an agent? Least privilege isn't an AI-specific concept — it's a...

intermediate

Prompt Injection Defense

Designing a mechanism so an agent, while processing external data (like webpage content or user-uploaded documents), can distinguish original task instructions from malicious instructions potentially hidden within the data itself, avoiding being misled into performing unintended actions by content buried in that data.

What is prompt injection defense, and how does it differ from typical <a href="/en/glossary/prompt-techniques/jailbreaking/">jailbreaking</a>...

intermediate

Weekly Picks

news

Barclays Scales Up Claude: Targeting 50% Developer Adoption of Claude Code by Year End, Sorting 120,000 Emails a Day

16,000 employees, 120,000 emails a day, a million searches logged — Barclays' disclosure isn't a...

Grace Patel

practice

Claude Agent SDK Adds verbatim_prompts: Turning Off Automatic @path Expansion and Slash-Command Dispatch to Stop Untrusted Text From Executing as Commands

A stray @/etc/passwd hidden in external input can trigger an unintended file read —...

Marcus Ide

practice

Claude Agent SDK Fixes a Background-Subagent Bug: stdin Closing Too Early Right When a Subagent Finishes Breaks the Next Turn

A background subagent finishing right as the main result arrives used to close the connection...

Derek Finch

practice

Claude Agent SDK Adds prewarm(): Starting a Claude Code Process Before the Session Is Even Known, to Cut First-Query Latency

Process-startup latency isn't unsolvable, it just needed someone to split the problem —...

Marcus Ide