0.10.0 (2026-05-20)
Full Changelog: v0.9.0...v0.10.0
Features
- add the pronunciation dict (665cf91)
- api: add auto-scale/schedule/OIDC params, SCHEDULED_CAPACITY billing to clusters (7769868)
- api: add cached_input field to ModelObjectPricing (fb7b530)
- api: add disable_position_bias_correction param, remove num_samples from compare results (a1908f0)
- api: add h100-40gb-mig, b200-192gb GPU types, remove a100-80gb from deployments (e041234)
- api: add h200-140gb GPU type option to betajig (fedd305)
- api: add instance_name field to Remediation (1f5e3cd)
- api: add max_seq_length field to finetuning (0964ef9)
- api: add max_tokens and temperature parameters to eval judge (c619446)
- api: Add node remediation APIs to clusters sdks (a44d364)
- api: add num_workers parameter to eval evaluation parameters (d061b9d)
- api: add parameters to clusters/storage list methods, update cluster/storage types (6e07f04)
- api: add storage/multipart/logs endpoints, update deployment/volume types (8178f61)
- api: add trigger filter and multi-mode support to cluster remediation list (e73e7ee)
- api: allow string or object for audio_inputs/source_video in video params (75b4202)
- api: Integrate fine_tuning.list_metrics from stainless (9b4d021)
- api: manual updates (4de5220)
- api: manual updates (4140f1d)
- api: Update server url to .ai (9ae586b)
- client: optimize json encoder for internal types (f35ac5f)
- ENG-87042: document language on TTS WebSocket and simplify locale note (454dc61)
- go: add default http client with timeout (f235d6f)
- MOSH-2181: Add default note on max-seq-length (ee06e5a)
- support setting headers via env (56a6ac9)
Bug Fixes
- api: remove trigger param from beta cluster remediation list (1249f1c)
- better respect format tags from the spec (a99fd75)
- client: update base URL to api.together.ai in video service (169e53b)
- go: avoid panic when http.DefaultTransport is wrapped (d3d1491)
- types: change language from enum to string in audio speech (c7c3dd5)
- types: change status to enum in ClusterStorage (0ae8433)
- types: convert endpoint to enum in batch new params (24f44d8)
- types: make duration_days and size_tib optional in beta clusters (5d5ba57)
- types: make RequestID required, remove error field in BetaJigQueueSubmitResponse (f395a88)
- types: remove eval-sample/eval-output/eval-summary/batch-generated from FilePurpose (27aa4f5)
- types: remove node_name from ClusterControlPlaneNode and ClusterGPUWorkerNode (150c1a4)
- types: remove task field from audiotranscription and audiotranslation responses (c0e0c08)
- types: rename cuda_driver_version to cuda_version in cluster regions (0d3c446)
- types: restructure driver_versions, make supported_instance_types required in betacluster (1298ca1)
Chores
- avoid embedding reflect.Type for dead code elimination (8f52477)
- internal: more robust bootstrap script (32acdb4)
- redact api-key headers in debug logs (8b4cdf7)
- temp deletion (deaf58b)
- temp undeletion (07426d5)
- tests: bump steady to v0.22.1 (998c407)
Documentation
- api: clarify gradient_accumulation_steps default behavior in finetuning (2d6b8ac)
- api: clarify max_seq_length parameter in finetuning (3148bba)
- api: clarify prompt parameter docs in transcription/translation (637afa2)
- api: clarify suffix parameter character limit in fine-tuning (6f8178d)
- api: document voice mixing support for audio speech voice parameter (374f51b)
- api: update billing_type description in beta cluster creation (a40ff7f)
- api: update docstrings from future to present tense (7795a05)
- api: update sample_rate and response_encoding parameter docs in audiospeech (e1b61d9)
- api: update supported formats in audio transcription/translation file params (3bf7862)
- improve examples (28ad2a1)
- types: add field descriptions to cluster/storage types (8773a13)