Anthropic reports a substantially cheaper, higher-scoring Haiku version, but the available evidence is a single-source account of a small-model update rather than an independently verified breakthrough.
Limits of the evidence: Discovery-class roundup; performance, pricing and availability claims are attributed to Anthropic or Microsoft, with no independent corroboration in the supplied text.
What to watch
Assessment revised Oct 15, 2026
Anthropic gives Claude Haiku a big update: Microsoft-Decision-1 responds to Jev and OpenAI
Andrew NgOct 12, 2026First report
First assessment, from Andrew Ng
No maturity ladder applies to this desk.
Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original sources are authoritative.