Releasesby IbisW&B Weave 0.53.12
Update11 days agoUnverifiedAdded Oct 17, 2026
- fix(weave): preserve complete usage across trace batches fix(weave): evaluate filtered call pages once perf(weave): bound trace attribution for single-span lookups fix(weave): preserve model tool calls in chat projection fix(weave): propagate span expiry to message data chore(weave): update model costs chore(weave): leave the vendored client alone for a...
Topics: AI agents, LLMs, SQL, Pricing, Observability
More W&B Weave releases
Every W&B Weave releaseW&B Weave 0.53.11
- fix(weave): order sibling agent spans sharing a start by earliest end chore(dev): remove codecov integration chore(dev): add git-backed Devin blueprint feat(weave): standardize SDK attribution under wandb.sdk fix(weave): cap opentelemetry-exporter-otlp-proto-http below 1.45
W&B Weave 0.53.10
- chore(weave): refresh vendored trace-server clients fix(weave): return 400 for an illegal query argument, not 403 fix(weave): unbox traced values before the google.genai call fix(weave): support current-version call queries feature(weave): make Stainless default HTTP client chore(weave): refresh vendored trace-server clients chore(weave): document the V2...
W&B Weave 0.53.9
Release version: 0.53.9-dev0 → 0.53.9
W&B Weave 0.53.8
Release version: 0.53.8-dev0 → 0.53.8
W&B Weave 0.53.7
- chore(weave_ts): add 0.16.6 changelog entry fix(weave): map ClickHouse stream failures to 502 and read structured error codes feat(weave): allow score tracing to be disabled in imperative evals fix(weave): keep eval rows that produced no score in the summary fix(weave_ts): report inclusive Anthropic prompt tokens feat(weave): add tables for tags...
W&B Weave 0.53.6
- fix(weave_ts): return the real image from client.get() fix(weave_ts): label a fetched image with its real format fix(weave): keep chat completion streams open for iteration feat(weave): trace Claude Agent SDK subagents fix(weave): trace Claude Agent SDK async-iterable queries per turn fix(weave): stop UnboundLocalError in the MCP read_resource wrapper...
W&B Weave 0.53.4
- fix(weave_ts): restore tracing for structured outputs on openai-node v5+ feat(weave): add custom runtime registration API feat(weave): record turn output messages fix(weave): preserve custom provider names in selectors chore(weave): ignore (and log) extra fields from BaseModelStrict refactor(weave_ts): use Weave GenAI SDK for Claude Agent tracing...
W&B Weave 0.53.3
- fix(weave): stop bool/int param collision that 400s calls queries chore(weave): file storage presign_read chore(weave): export target queries fix(weave): preserve object read error context chore(inference): use uv and add sync_model_catalog target chore(weave): clickhouse export write path fix(weave): map openai cached_tokens to cache_read on otel ingest...
Also shipped on Oct 13, 2026
CoreWeave in October 2026What's included:
- Three training methods: Supervised fine-tuning (SFT) , direct preference optimization (DPO) , and reinforcement learning with group relative policy optimization (GRPO) . Each method trains a low-rank adapter (LoRA) or the full model weights. Custom loss functions : Train with any objective you compute in PyTorch from the model's log-probabilities.
The AI.AGG function is generally available (GA)
The AI.AGG function is generally available (GA). You can use this function to semantically aggregate unstructured input data based on natural language instructions.
BigQuery generative AI functions now support the gemini-4-argon model
BigQuery generative AI functions now support the gemini-4-argon model. This model is generally available .
Gemini 4 Argon in CodeMender (Preview)
CodeMender now supports Gemini 4 Argon. To use it, pass --model gemini-4-argon to cm find , cm verify , or cm fix . Gemini 3.8 Flash remains the default model. For more information, see Specifying the model .
Gemini 4 Argon and Gemini 4 Argon Cyber Permissive
Gemini 4 Argon ( gemini-4-argon ) is available in General Availability (GA) . Gemini 4 Argon is Google's frontier model that excels at deep work in complex, real-world tasks across domains including coding, finance, legal, cybersecurity, and science.
On-demand state repartitioning is now generally available
On-demand state repartitioning is now generally available. You can resize the number of shuffle partitions for a stateful Structured Streaming query or Lakeflow pipeline without losing checkpoint state. See On-demand state repartitioning for stateful streaming queries .