09.08.2026: New Model Providers, Trace Filters, and PXI Updates

Run Z.ai, MiniMax, and Meta models in the playground and PXI, filter traces by error status and latency from both SDKs, and find PXI settings and slash commands faster.

Z.ai, MiniMax, and Meta join the built-in model providers, and both SDKs can now ask for just the slow failures.

Z.ai, MiniMax, and Meta as Built-In Providers

September 2 – 8, 2026

Available in arize-phoenix 20.6.0+ (Z.ai), 20.8.0+ (MiniMax), and 20.9.0+ (Meta); provider constants in @arizeai/phoenix-client 7.8.0+ and @arizeai/phoenix-cli 1.17.0+

Three providers that previously needed a custom OpenAI-compatible endpoint are now built in. Pick them in the playground, save them on prompts, and use them for PXI sessions. Token prices ship with them, so cost tracking works without manual setup.

Provider	Models	Credential	Base URL override
Z.ai	glm-4.6, glm-4.5, glm-4.5-air	ZAI_API_KEY	ZAI_BASE_URL
MiniMax	MiniMax-M3, MiniMax-M2.7	MINIMAX_API_KEY	MINIMAX_BASE_URL
Meta	muse-spark-1.3, muse-spark-1.3-contributor, muse-spark-1.2, muse-spark-1.2-contributor, muse-spark-1.1	META_API_KEY	META_BASE_URL

• Self-hosted deployments can gate them with ZAI, MINIMAX, and META in PHOENIX_ALLOWED_PROVIDERS.
• New models on existing providers: gemini-3.8-flash on Google (20.6.0+) and gpt-6-astra on OpenAI (20.8.0+), with reasoning controls and pricing.

Prompt Playground

Compare models side by side

Custom AI Providers

Point a provider at your own gateway

Filter Traces by Error Status and Latency

September 3 – 8, 2026

Available in arize-phoenix 20.8.0+ (server), arize-phoenix-client 3.5.0+ (Python), and @arizeai/phoenix-client 7.10.0+ (TypeScript)

Both SDKs now take error, min_latency_ms, and max_latency_ms when listing traces, so “which traces errored and took more than a second?” is one call instead of a page of traces filtered in your own code.

• error cuts both ways. true returns traces with at least one ERROR span, matching the UI indicator. false returns only clean traces, which is what you want when sampling healthy behavior for a dataset.
• Latency bounds are inclusive and apply to the trace, not its spans.
• The SDKs fail fast. An older server reports an upgrade requirement instead of silently returning unfiltered traces, and inverted bounds raise client-side.

from phoenix.client import Client

client = Client()

slow_failures = client.traces.get_traces(
    project_identifier="my-project",
    error=True,
    min_latency_ms=1000,
    limit=50,
)

import { getTraces } from "@arizeai/phoenix-client/traces";

const slowFailures = await getTraces({
  project: { projectName: "my-project" },
  error: true,
  minLatencyMs: 1000,
  limit: 50,
});

Filter Expressions

Filter spans, traces, and sessions with the query DSL

List Traces for a Project

The full set of query parameters

PXI: Tabbed Settings and Slash Command Suggestions

September 2 – 8, 2026

Available in arize-phoenix 20.6.0+ (settings) and 20.9.0+ (suggestions)

• Assistant settings are five linkable tabs — General, Tools, Permissions, Tracing & privacy, and Chats & data — each with its own URL under /settings/agents. Settings for capabilities your deployment has turned off stay hidden.
• The chat input suggests a slash command that fits what you have in context. Open a chat with a span attached and the placeholder points at the command for that situation.

PXI

Configure the AI engineering agent built into Phoenix

MCP Server on FastMCP 4

September 3, 2026

Available in arize-phoenix 20.7.0+

The mounted MCP server at /mcp moves to FastMCP 4 and the stateless protocol, which removes a class of reconnect problems. Nothing changes for clients on the default code-mode surface.

• Progressive disclosure is gone. list_tool_groups and enable_tool_group no longer exist. With PHOENIX_ENABLE_MCP_CODE_MODE=false, every /v1 operation is advertised as its own tool.
• Analytics SQL errors read properly, so a bad query comes back as an actionable message.

Remote MCP Server

Connect an MCP client to Phoenix

Delete Traces from a Project over REST

September 2, 2026

Available in arize-phoenix 20.6.0+

DELETE /v1/projects/{project_identifier}/traces clears traces from a project while keeping the project and its configuration, so ingestion continues against the same project afterward.

• The time window is required. start_time and end_time bound the deletion as [start_time, end_time). Naive datetimes are read as UTC.
• Spans and orphaned sessions go with them. Sessions that still have traces outside the window survive.

curl -X DELETE \
  "$PHOENIX_ENDPOINT/v1/projects/my-project/traces?start_time=2026-08-01T00:00:00Z&end_time=2026-09-01T00:00:00Z" \
  -H "Authorization: Bearer $PHOENIX_API_KEY"

Delete Traces from a Project

Endpoint reference

Data Retention

Expire traces on a schedule instead

Additional Improvements

September 2 – 8, 2026

Available in arize-phoenix 20.6.0+ and arize-phoenix-client 3.5.0+

• Project token totals stop double-counting. Only leaf LLM spans count now, so frameworks that nest an LLM span inside another no longer inflate the dashboard headline.
• Metadata cells filter the table you are on. Clicking a metadata value in the traces table appends to the traces filter, not the span filter.
• Faster filtered trace and session queries. Filter lowering correlates every quantifier, which on PostgreSQL turns a statement timeout into an index probe.
• Annotation uploads validate every ID column up front, so an invalid document_position fails with a clear error instead of a confusing one later.
• PXI stops retrying doomed GitHub calls. A 401 or 403 is terminal and reports that the token’s scope or repository access needs checking.
• Experiments no longer finish early. A dispatched work item counts as in-flight from the moment it leaves the queue.
• Tool definitions carry their names on PXI’s own LLM spans, so its traces read cleanly in span details.
• Question counts hold steady while a PXI question streams in.