NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI · #4785 most downloaded on PyPI
AI Observability and Evaluation
Last release 4 days ago
01 Oct 2026
Ships on a steady schedule
a new release about every 8 days
Nearly every release is documented
notes for 60 of the last 60 stable releases
19 versions withdrawn
withdrawn after publishing
4 years old
700 releases · first in 2023
One column per quarter.
use DISTINCT instead of GROUP BY in ProjectHasTracesDataLoader
rest-api: DELETE /prompt_versions/{id}/tags/{tag_name}
ui: fix model menu search losing focus on first keystroke
TracesTable: update interface for additional spans row and adjust GraphQL fragments
add PromptInput compound component for AI interactions
deps: update arize-phoenix-evals to 2.12.0
Pin LiteLLM <1.82.7 to mitigate supply chain attack
deprecate useNotifyError and replace error toasts with inline Alert banners
px docs fetch (#12162) (c0836c3)Add phoenix db migrate subcommand for standalone migrations
phoenix db migrate subcommand for standalone migrations (#12071) (11360a6)add phoenix-pr-screenshot slash command skill
client: remove deprecated client.annotations module
client.annotations module (#11983) (6203c31)Add feature-flagged tracing onboarding empty state
Phoenix now supports Perplexity AI and Together AI as built-in providers in the playground. Both use OpenAI-compatible APIs — just set your API key an
Phoenix now supports Perplexity AI and Together AI as built-in providers in the playground. Both use OpenAI-compatible APIs — just set your API key and start chatting.
| Provider | API Key Env Var | Base URL |
|---|---|---|
| Perplexity AI | PERPLEXITY_API_KEY |
https://api.perplexity.ai |
| Together AI | TOGETHER_API_KEY |
https://api.together.xyz/v1 |
Perplexity's Sonar model family — search-augmented LLMs that ground responses in real-time web results — is now available directly in the playground.
moonshotai/Kimi-K2.5deepseek-ai/DeepSeek-V3.1openai/gpt-oss-120bopenai/gpt-oss-20bzai-org/GLM-5zai-org/GLM-4.5-Air-FP8Qwen/Qwen3-235B-A22B-Thinking-2507deepseek-ai/DeepSeek-R1Cost tracking is included out of the box with 9 Perplexity and 26 Together AI model entries added to the cost manifest.
pip install 'arize-phoenix>=13.11.0'
docker pull arizephoenix/phoenix:13.11.0
Full Changelog: https://github.com/Arize-ai/phoenix/compare/arize-phoenix-v13.10.0...arize-phoenix-v13.11.0
Phoenix now supports Cerebras, Fireworks AI, Groq, and Moonshot (Kimi) as first-class providers in the playground. All four use OpenAI-compatible APIs
Phoenix now supports Cerebras, Fireworks AI, Groq, and Moonshot (Kimi) as first-class providers in the playground. All four use OpenAI-compatible APIs — no new dependencies required. Just set the appropriate API key environment variable and start chatting.
| Provider | API Key Env Var |
|---|---|
| Cerebras | CEREBRAS_API_KEY |
| Fireworks AI | FIREWORKS_API_KEY |
| Groq | GROQ_API_KEY |
| Moonshot (Kimi) | MOONSHOT_API_KEY |
Cost tracking is also included out of the box with 298 new model entries across all four providers, so token costs are calculated automatically.
You can now edit a project's description and gradient colors directly from the Project Settings tab. A new patchProject GraphQL mutation powers this on the backend.
The playground model list now includes the latest OpenAI reasoning models: gpt-5.4 family, gpt-5.3-chat-latest, gpt-5.2-pro variants, gpt-5 date-stamped and pro variants, and o3-pro-2025-06-10.
pip install 'arize-phoenix>=13.10.0'
docker pull arizephoenix/phoenix:13.10.0
Full Changelog: https://github.com/Arize-ai/phoenix/compare/arize-phoenix-v13.9.0...arize-phoenix-v13.10.0
api: add trace_id filter to GET spans REST endpoints
add integration icons for claude, langgraph
js: bump minimatch to >=10.2.3 to fix ReDoS vulnerabilities
add configurable password policy
add nav counters for projects, datasets, evaluators, and prompts
evals: add conciseness classification evaluator
add AWS Bedrock cross-region inference model prefix preference
use store-and-network fetch policy for ModelMenu query
Enhance table functionality with column resizing support in EvaluatorsTable and DatasetEvaluatorsTable
Prevent playground rerun when opening trace slide-over
deps: update arize-phoenix-client to 1.29.0
add Claude Opus 4.6 model to playground
graphql: access ORM attributes directly in asExampleRevision
oauth2: add EMAIL_ATTRIBUTE_PATH for configurable email extraction
add tool invocation accuracy metric
deps: update arize-phoenix-client to 1.28.0
add FaithfulnessEvaluator and deprecate HallucinationEvaluator
Add AttributesJSONBlock component for enhanced JSON attribute display
Add connection timeout error handling
handle non-array values in NumDocuments to prevent PostgreSQL jsonb_array_length error
add support for JSONL dataset uploads
add generate api key buttons to project onboarding modals
move add dataset example to be a part of the examples tab
cost: update built-in model token prices
Lazy load embeddings/dimension features
null as sentinel value for PHOENIX_LDAP_ATTR_EMAIL (#10560) (2e121cd)## 12.23.0 (2025-12-10) ### Features * make email optional for LDAP (#10525) (916adb0) ### Documentation * december release notes
Lightweight Directory Access Protocol (LDAP)
cost: update built-in model token prices
Add support for claude opus 4-5
Improve UX of retention policy form
implement connection pagination on trace spans
cost: update built-in model token prices
## 12.15.0 (2025-11-13) ### Features * Update Anthropic model list
## 12.14.1 (2025-11-13) ### Bug Fixes * add type safety to loader data
Add react compiler behind a config
Your coding agent can read these notes before it upgrades. Set up the MCP server →