Claude Reportedly Injects Massive Context Server-Side
MaleficentSurprise · reddit · 2026-07-19
Reddit users report Claude throwing errors almost daily in new conversations: This request exceeds Claude's context limit by about 4-6M tokens, affecting web, desktop, iOS, and incognito modes.
Further investigation reveals the error isn't triggered by user input, but rather by the server injecting a 19MB / 4.8M token object during prompt assembly; different models behave identically, independent of the payload. Worse, on usage-based billing accounts, this injected content is actually processed and billed if no error occurs or if it's under the context limit, potentially leading to $4 charges per message. The post also notes a lack of support response, and the author seeks to collect affected account dates, excess tokens, surfaces, and billing types to pinpoint the issue.
More from Models
- Unverified Grok Bot concept: connect banks and brokers as your personal CFO — nima_owji · 2026-10-26
- NVIDIA's Nemotron 3.5 Super VL scores just 1.94% on ARC-AGI-2 — teortaxesTex · 2026-10-26
- Tencent's Hy-SimulMT lands on Hugging Face in six variants — AdinaYakup · 2026-10-26
- Fintech insiders' AI consensus: ChatGPT, Claude and Grokbot top the pack — LexSokolin · 2026-10-26
- DeepSeek-adjacent researcher hints next model will keep the V3.2 DSA approach — YouJiacheng · 2026-10-26
- Anthropic S-1: 1,300% YoY revenue growth, 85% gross margin — danshipper · 2026-10-26