Honeycomb vs New Relic for LLM tracing

Here’s the short version:

• Honeycomb is usually the better fit if you want deep, high-cardinality, query-first tracing/observability for LLM apps and you expect to inspect lots of attributes like prompts, model names, tenant IDs, retrieval metadata, tool calls, token usage, etc.
• New Relic is usually better if you want a broader all-in-one observability platform that already covers infra, logs, APM, and you want LLM tracing to live inside that existing stack.

How they differ for LLM tracing

Honeycomb

Strengths

• Excellent for event-level analysis and slicing traces by many dimensions.
• Handles high-cardinality data very well, which is useful in LLM systems.
• Great when you need to answer questions like:
  • Which prompt template caused latency spikes?
  • Which customer/tenant is hitting tool failures?
  • Which retrieval source correlates with bad outputs?
  • How do token counts vary by route/model/version?
• Strong for exploratory debugging and building custom workflows around traces.

Tradeoffs

• Not as “suite-like” if you also want logs, infra, dashboards, alerting, error tracking all in one place.
• Requires a bit more intentional modeling of telemetry to get the most value.

New Relic

Strengths

• Broad platform: APM, infrastructure, logs, browser/mobile, errors, synthetics, plus LLM observability features.
• Easier if your team already uses New Relic for everything else.
• Good for correlating LLM traces with the rest of the app stack in one place.

Tradeoffs

• Can feel heavier and less “investigative” than Honeycomb for very deep trace slicing.
• High-cardinality workflows are often where Honeycomb tends to shine more naturally.

For LLM-specific use cases

If you care about things like:

• prompt/version comparisons
• per-tenant or per-user analysis
• RAG retrieval metadata
• tool/function call tracing
• token usage and cost
• latency broken down by model/provider/retriever/tool

…then Honeycomb often feels more natural for analysis.

If your goal is:

• “I want all observability in one vendor”
• “My SRE team already lives in New Relic”
• “I want LLM traces alongside infra, logs, and errors without much setup”

…then New Relic may be the simpler choice.

Practical recommendation

• Choose Honeycomb if your main pain is debugging and understanding LLM behavior.
• Choose New Relic if your main pain is platform consolidation and you already rely on New Relic broadly.

Rule of thumb

• LLM product teams / platform teams: Honeycomb
• General observability teams wanting one tool: New Relic

If you want, I can also give you:

1. a feature-by-feature comparison table,
2. a pricing/effort tradeoff view, or
3. a recommended setup for OpenTelemetry-based LLM tracing in either tool.