A million tokens changes the math, not the magic
AI Changelog · Issue 11Week 23 · Jun 1–7, 2026
Claude 4.6 Opus shipped a 1M-token context window and everyone declared retrieval dead. It isn't, but our long-document verdict did split for the first time, and the deciding axis is now corpus size, not model capability.
52 announcements scanned · 6 mattered · 1 verdict changed
Verdicts changed
Best approach for long-document analysis (<50 docs)
Claude Opus 4.8
Retrieval pipeline (RAG)previous pick
Single-pass Claude 4.6 Opusnew pick
The 1M context window removes chunking overhead below ~50 documents. RAG still wins at volume.
On our radar
Claude 4.6 Opus agentic coding claims
Benchmarking against our current pick over the next two weeks of pipeline runs.
Filtered out
- Three "context is all you need" thinkpieces: capability claims without workload data.
- 11 wrapper tools repackaging the 1M context window as a product.
Our take: Bigger context windows don't kill retrieval: they move the break-even point. Know your corpus size and the answer falls out.
Neomanex, from our own production runs
Get issue 12
Next week's verdict changes and deadlines, in your inbox.