Weights & Biases 0.31.2
Weights & Biases 0.31.2 introduces EvalTable backed by the CoreWeave Evaluation Service in public preview and includes fixes for sweep scheduler parameters.
Also shipped Nov 21, 2026CoreWeave in November 2026 →
[Launched] Generally Available: Governance policy guidance in Foundry Canvas and VS Code
Microsoft Foundry governance policy guidance is now generally available in Foundry Canvas and the Foundry Toolkit for VS Code, surfacing relevant policies for agents and models.
[In preview] Public Preview: Microsoft Copilot harness support in Foundry Toolkit for VS Code
The Foundry Toolkit for VS Code now supports prompt agents using the Microsoft Copilot harness in public preview, allowing developers to configure tools and deploy changes.
antfly 0.3.2
Antfly v0.3.2 introduces cached Gemma cooperative decoding, worker placement hints, and various performance optimizations for inference and semantic search.
Web Search on Amazon Bedrock expands language support and country-aware search
Web Search on Amazon Bedrock now supports five additional languages and an optional country field to bias search results toward specific geographic locations.
Qwen Code Release v0.28.3
Qwen Code v0.28.3 adds a unified chat surface contract, GitHub as a managed channel adapter, and a generic webhook channel for notifications and task results.
More from Weights & BiasesEvery Weights & Biases release →
Weights & Biases 0.31.1
Weights & Biases 0.31.1 adds a backup sync function for offline runs, sandbox event support in the automations API, and new features for the W&B LEET TUI.
Weights & Biases 0.31.0
Notable Changes - The default API endpoint is now CoreWeave Forge at https://api.forge.coreweave.com, and run links point to https://forge.coreweave.com/wandb. Existing API keys stored for api.wandb.ai keep working without logging in again. Added - Python 3.15 is now supported (@dmitryduev in ) - wandb.init(resume="must") now works for offline runs.
Weights & Biases 0.30.0
Notable Changes wandb.InternalApi , the wandb.PublicApi and wandb.apis.PublicApi aliases, and the wandb.apis.internal and wandb.sdk.internal.internal_api modules have been removed. Use wandb.Api() instead. wandb.api and wandb.ensure_configured() are deprecated.
Weights & Biases 0.29.0
Notable Changes Removed the deprecated start_method setting that already had no effect. The .length property on paginated objects (e.g. Api.runs() , Api.sweeps() , run.files() ) has been removed. Use len(...) instead. wandb.Video() now requires the format argument when initializing from a numpy array or raw bytes.
Weights & Biases 0.28.2
Notable Changes The wandb sync --clean command now exits with code 1 and prints a hint to use wandb clean , which is the replacement. wandb login now verifies credentials by default. This can be disabled with wandb login --no-verify or programmatically with wandb.login(verify=False) .
Weights & Biases 0.28.1
Notable Changes This version drops support for protobuf v4, and requires protobuf v5 or newer. This version drops compatibility with server versions older than 0.70.0.
Weights & Biases 0.28.0
Notable Changes This version drops compatibility with server versions older than 0.65.0. Added High-resolution image rendering in terminals supporting the Kitty protocol with ANSI fallback in the W&B LEET TUI media pane ( wandb leet command) ( @dmitryduev in #11806 ) Synced scrubbing in the W&B LEET media pane: press l to link scrubbing, then the scrub keys...
Weights & Biases 0.27.2
Fixed Run.scan_history() no longer returns duplicate or missing steps when getting run history ( @jacobromero in #12010 )