Datadog

Experiment result retention now extends from 15 days by default to 3 months when customers have an annual or month-to-month commitment. Extended data retention add-ons for LLM Observability increased across all tiers: 30-day retention $1.50->$6.50, 60-day $3->$8, 90-day $4->$9 per 10K LLM spans/month.

Product renamed 'LLM Observability' -> 'Agent Observability' and pricing model overhauled: replaced single starting tier ($8/10K monitored LLM requests/month, $12 on-demand, min 100K requests/month) with a Free tier ($0, up to 40K LLM spans/month) plus a Pro tier ($160/mo annual, $200/mo month-to-month, $240/mo on-demand for 100K LLM spans, then $3.50-$5.00 per additional 10K spans). New paid data-retention add-ons introduced: 30-day ($1.50/10K spans/mo), 60-day ($3/10K spans/mo), 90-day ($4/10K spans/mo) trace retention.

Datadog renamed its LLM Observability product to Agent Observability. All pricing ($0 Free, $160/mo Pro), plans, and features remain unchanged — only the product name changed.

Pro plan card now displays overage pricing explicitly: 'Additional spans: $3.5 per 10K LLM spans.' Span inclusion bullet updated to '100K LLM spans per month included'. FAQ expanded with 3 overage tiers: $3.5 annual, $4.20 M2M, $5.00 on-demand. Pro plan tagline updated from 'For teams running in production' to 'For teams at scale', broadening the described audience.

Free tier added: $0/month for individuals/small teams with up to 40K LLM spans/month, 15-day trace retention, unlimited context and evaluations, and full feature access. No credit card required. LLM Observability relaunched with Free/Pro tiers replacing per-request pricing. Free: $0/mo, 40K spans. Pro: $160/mo annual ($200 M2M, $240 on-demand) for 100K LLM spans. Billing unit changed to LLM spans.

Datadog added a notice announcing new pricing for LLM Observability effective May 1, 2026, with a link to more information. Current pricing ($8/10K requests annually, $12 on-demand) unchanged.

New Experiments feature: Debug traces in Playground by testing prompts, parameters, and models New Experiments feature: Select best configurations with side-by-side experiment comparisons