phoenix

varize-phoenix-v16.4.0 scope: arize-phoenix Feature

This release adds 3 notable features for engineering teams evaluating rollout.

Published 2d Tracing

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agents ai-monitoring ai-observability aiengineering anthropic datasets

+10 more

evals langchain llamaindex llm-eval llm-evaluation llmops llms openai prompt-engineering smolagents

Summary

AI summary

Updates 16.4.0, Bug Fixes, and 2026-06-01 across a mixed release.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Medium	Add no-data chart overlays Add no-data chart overlays Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Collapse agent context pills into a deck Collapse agent context pills into a deck Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Make PXI quick actions dynamic per page context Make PXI quick actions dynamic per page context Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Add Claude Opus 4.8 support to PXI agents Add Claude Opus 4.8 support to PXI agents Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Add span annotation tool to agents Add span annotation tool to agents Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Improve floating chat drag, resize, and style in agents Improve floating chat drag, resize, and style in agents Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Show Beta badge in PXI header on new chats Show Beta badge in PXI header on new chats Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Add trace debugging skill to agents Add trace debugging skill to agents Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Medium	Upgrade eval harness with message history in agents Upgrade eval harness with message history in agents Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Low	Style code blocks with the Pierre theme for consistent light/dark Style code blocks with the Pierre theme for consistent light/dark Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Feature	Low	Make span_kind comparisons case-insensitive in filter Make span_kind comparisons case-insensitive in filter Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Feature	Low	Lift trace UX feature flag Lift trace UX feature flag Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Feature	Low	Add playground skill Add playground skill Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Feature	Low	Surface PXI model credential setup UI Surface PXI model credential setup UI Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix
Bugfix	Medium	Avoid PXI initial scroll animation Avoid PXI initial scroll animation Source: llm_adapter@2026-06-01 Confidence: high	—
Bugfix	Medium	Update built-in model token prices in cost module Update built-in model token prices in cost module Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Adjust spans sparkline bar gap Adjust spans sparkline bar gap Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Keep PXI thinking indicator after tool calls Keep PXI thinking indicator after tool calls Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Load evaluator subclass fields in dataloader Load evaluator subclass fields in dataloader Source: granite4.1:30b@2026-06-01-audit Confidence: low	—

Full changelog

16.4.0 (2026-06-01)

Features

Add no-data chart overlays (#13471) (792aa92)
agent: collapse agent context pills into a deck (#13516) (f1e0e2e)
agent: make PXI quick actions dynamic per page context (#13534) (e60ba32)
agents: add Claude Opus 4.8 support to PXI (#13497) (f13fd0b)
agents: add span annotation tool (#13502) (4c8262c)
agents: Floating chat drag, resize, style improvements (#13443) (79d8afe)
agent: show a Beta badge in the PXI header on new chats (#13535) (1deb767)
agents: trace debugging skill (#13467) (7ddc699)
agents: upgrade eval harness with message history (#13474) (b8cc702)
app: style code blocks with the Pierre theme for consistent light/dark (#13519) (a3ec631)
filter: make span_kind comparisons case-insensitive (#13489) (a1970a3)
lift trace UX feature flag (#13470) (f22e85b)
playground skill (#13465) (8ec9392)
surface PXI model credential setup (#13458) (566a1f5)

Bug Fixes

adjust spans sparkline bar gap (#13500) (8da7337)
avoid PXI initial scroll animation (#13456) (6a408ad)
cost: update built-in model token prices (#13522) (d14a55c)
keep PXI thinking indicator after tool calls (#13488) (dfd0ec6)
load evaluator subclass fields in dataloader (#13449) (1a36841)

Documentation

Add Phoenix release notes for 05-21-2026 through 05-27-2026 (#13446) (6200ae2)
add ReleaseUpdate component with read more links (#13544) (ffa4762)
agents: note python and node are unavailable in PXI bash tool (#13493) (1718c71)
graphql: describe Span token count fields and TokenCountPromptDetails (#13491) (a2ffe12)
introduce PXI agent (#13476) (c5fe607)
skills: weekly audit — 2026-05-27 (3108194)

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track phoenix

Get notified when new releases ship.

About phoenix

AI Observability & Evaluation

All releases →

Related context

Related tools

Earlier breaking changes

varize-phoenix-v17.0.0 Adds system settings for admin-managed assistant enablement and trace recording policy
varize-phoenix-v16.0.0 Sandboxing and Code Evaluators introduce breaking changes in Phoenix v16.0.0.
varize-phoenix-v15.7.0 Removes v1 /chat route and associated code