---
title: "What AI recommended for llm observability and evals in September 2026"
description: "What AI recommends for llm observability and evals, updated monthly: which tools Claude · Claude Haiku 4.5, Claude · Claude Opus 4.8, Claude · Claude Sonnet 4.6, ChatGPT, Venice · DeepSeek V4 Flash 0731, Venice · Gemini 3.6 Flash, Venice · Kimi K2.6, Venice · Qwen 3.7 Plus and Venice · GLM 5.2 named in September 2026 when developers asked \"What should I use for LLM observability and evals?\", cross-referenced with what developers say about them."
source: https://backtalk.sh/ai-recommends/llm-observability/2026-09
---

# What AI recommends for llm observability and evals (September 2026)

When developers ask "What should I use for LLM observability and evals?", this is what Claude · Claude Haiku 4.5, Claude · Claude Opus 4.8, Claude · Claude Sonnet 4.6, ChatGPT, Venice · DeepSeek V4 Flash 0731, Venice · Gemini 3.6 Flash, Venice · Kimi K2.6, Venice · Qwen 3.7 Plus and Venice · GLM 5.2 answer, measured over 445 sampled answers across 4 probe runs. Share of voice is the percentage of sampled answers naming the tool; the dev pulse is the sentiment of the same month's developer mentions on Hacker News, Reddit, GitHub, Stack Overflow, Bluesky and more.

| # | Tool | AI share of voice | Change vs prior month | Named by | Dev pulse |
|---|------|------------------:|----------------------:|----------|-----------|
| 1 | Arize Phoenix | 0% | -14.2 pts | none this month | not tracked |
| 2 | Braintrust | 0% | -15.9 pts | none this month | not tracked |
| 3 | DeepEval | 0% | -9.7 pts | none this month | not tracked |
| 4 | Helicone | 0% | -19.6 pts | none this month | not tracked |
| 5 | Langfuse | 0% | -22.8 pts | none this month | quiet (net +0) |
| 6 | LangSmith | 0% | -23.5 pts | none this month | not tracked |

Biggest mover: LangSmith (-23.5 pts vs the previous month).

## The prompts behind these standings

- What should I use for LLM observability? (recommendation)
- What are the best LangSmith alternatives? (alternatives)
- Which tool is best for evaluating LLM outputs? (recommendation)

## Method

These September 2026 standings are final and never change; cite them by permalink. Full methodology: https://backtalk.sh/ai-recommends/methodology (as markdown: https://backtalk.sh/ai-recommends/methodology.md). Free to cite with attribution to "BackTalk AI Recommendation Standings" and a link.