// AI Platform Drift

Claude Haiku 5.5 breaks budget_tokens parameter, enables thinking by default

Claude Platform release notes · published 2026-10-07 23:07 UTC
ACTION NEEDED MIGRATE MODELAPIDEPRECATION
WHAT CHANGEDClaude Haiku 5.5 (claude-haiku-5-5) launched October 7, 2026 with 1M token context window and 128k max output. The budget_tokens parameter now returns a 400 error. Adaptive thinking is enabled by default, so responses can begin with thinking blocks. The same text counts as more tokens than in Claude Haiku 4.5.
WHY IT MATTERSExisting code calling claude-haiku-4-5 with budget_tokens will fail with 400 errors if switched to claude-haiku-5-5. Response parsing that expects content to start immediately will break when thinking blocks appear first. Token budgets calculated for 4.5 will underestimate 5.5 costs for identical inputs.
WHAT TO DOGrep your codebase for budget_tokens and remove it before upgrading to claude-haiku-5-5. Update response parsers to handle thinking blocks at the start of responses. Review the migration guide at platform.claude.com and retest token counting against your actual prompts to adjust cost forecasts.
Model-reported confidence 95% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## Claude Platform release notes — October 7, 2026
+https://platform.claude.com/docs/en/release-notes/overview#october-7-2026-4
+Wed, 07 Oct 2026 00:00:00 GMT
+
+We've launched Claude Haiku 5.5 (claude-haiku-5-5), our most capable model tuned for high-volume and latency-sensitive work. It has a 1M token context window, 128k max output tokens, and adaptive thinking with the effort parameter. It's available on the Claude API, Claude in Amazon Bedrock, Claude Platform on AWS, Claude on Google Cloud, and Claude in Microsoft Foundry. See What's new in Claude Haiku 5.5.
+Code written for Claude Haiku 4.5 can break on Claude Haiku 5.5. Manual extended thinking (budget_tokens) returns a 400 error, and adaptive thinking is on by default, so a response can begin with thinking blocks. The same text also counts as more tokens. See the migration guide. For model-specific prompting patterns, see Prompting Claude Haiku 5.5.
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.95, needed 0.45
triage reasoningNew model release with breaking changes and token accounting shifts affecting production workloads.
signalsnew model released (claude-haiku-5-5) · context window specified (1M tokens) · max output tokens specified (128k) · new feature (adaptive thinking with effort parameter) · breaking API change (manual extended thinking budget_tokens returns 400 error) · behavioral change (adaptive thinking on by default) · token counting change (same text counts as more tokens) · migration guide referenced
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)