// AI Platform Drift

Meta Llama 4 Maverick context window halved to 524K tokens

OpenRouter model catalogue · published 2026-10-19 18:44 UTC
ACTION NEEDED VERIFY_ONLY LIMITSMODEL
WHAT CHANGEDOpenRouter's catalogue now shows Meta Llama 4 Maverick with a context_length of 524288 tokens, down from 1048576 tokens. Pricing remains $0.1875 per 1M prompt tokens and $0.6525 per 1M completion tokens. Max completion tokens stay at 16384.
WHY IT MATTERSPrompts or documents exceeding 524288 tokens will now fail or be truncated. Any workflow relying on the full 1M token window—long document analysis, large codebases, multi-turn conversations with extensive history—will break.
WHAT TO DOAudit calls to meta-llama/llama-4-maverick for input token counts. If any exceed 524288, split the workload, switch models, or reduce context before the next deployment.
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
 name: Meta: Llama 4 Maverick
 pricing.prompt: 0.0000001875
 pricing.completion: 0.0000006525
-context_length: 1048576
+context_length: 524288
 top_provider.max_completion_tokens: 16384
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 0.5
triage reasoningContext window halved. Teams building on this model must verify compatibility and may need architectural changes.
signalscontext_length decreased from 1048576 to 524288 tokens (50% reduction) · model capability constraint change affecting prompt handling
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)