NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
npm · #3484 most downloaded on npm
Last release 2 days ago
01 Oct 2026
Ships on a steady schedule
a new release about every 8 days
Some releases are documented
notes for 32 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
2 years old
740 releases · first in 2024
Nothing published for this version
Updated dependencies [ 0343bb1 ]
Nothing published for this version
One column per month.
a105059 : fix(workflow): support deferred tool discovery in WorkflowAgent
Updated dependencies [ 4775577 ]
Nothing published for this version
Nothing published for this version
Nothing published for this version
1f29230: feat(mcp): harden oauth client registration according to latest protocol
1011e33: Select MCP OAuth authorization scopes from WWW-Authenticate challenges or Protected Resource Metadata.
WWW-Authenticate challenges or Protected Resource Metadata.@ai-sdk/openai-compatible@3.0.31
Updated dependencies [ 615ac89 ]
Updated dependencies [ b74971f ]
3b9f025: Add a provider-native reasoningEffort option for Open Responses endpoints.
reasoningEffort option for Open Responses endpoints.2338c79: feat (provider/openai): add jsdoc for openai tools
5819aec: fix (provider/openai): only send tool calls finish reason for tools that are not provider-executed
Updated dependencies [ a4ba394 ]
3204ab6 : fix(workflow): preserve per-tool types in execution callbacks and export the callback event types
experimental_transcribe fails with valid Buffera9a61b7: Add serviceTier to provider metadata for OpenAI responses
Updated dependencies [ 802af1e ]
180ce77 : fix(workflow): preserve tool behavior across workflow step boundaries
d18856a: fix(provider/openai): support websearch tool results without query property
15271d6: fix(provider/openai): do not set response_format to verbose_json if model is gpt-4o-transcribe or gpt-4o-mini-transcribe
These two models do not support it: https://platform.openai.com/docs/api-reference/audio/createTranscription#audio_createtranscription-response_format
Updated dependencies [1b5a3d3]
974de40: fix(provider/ai): do not set .providerMetadata.openai.logprobs to an array of empty arrays when using streamText()
.providerMetadata.openai.logprobs to an array of empty arrays when using streamText()### Patch Changes - Updated dependencies [0857788] - @ai-sdk/provider-utils@3.0.5
5e47d00: Support Responses API input_file file_url passthrough for PDFs.
5e47d00: Support Responses API input_file file_url passthrough for PDFs.
This adds:
Updated dependencies [ 850d863 ]
11109ae : feat(ai): support signed tool approvals in WorkflowAgent
a4bef93: feat(provider/openai): expose web search queries in responses api
a4bef93: feat(provider/openai): expose web search queries in responses api
6ed34cb: refactor(openai): consolidate model config into getResponsesModelConfig()
https://github.com/vercel/ai/pull/8038
Updated dependencies [ 6669d69 ]
ddc9d99: Implements logprobs for OpenAI providerOptions and providerMetaData in OpenAIResponsesLanguageModel
ddc9d99: Implements logprobs for OpenAI providerOptions and providerMetaData in OpenAIResponsesLanguageModel
You can now set providerOptions.openai.logprobs when using generateText() and retrieve logprobs from the response via result.providerMetadata?.openai
ec336a1: feat(provider/openai): add response_format to be supported by default
8d3c747: chore(openai): remove deprecated GPT-4.5-preview models and improve autocomplete control
097b452: feat(openai, azure): add configurable file ID prefixes for Responses API
fileIdPrefixes option to OpenAI Responses API configurationassistant- prefixed file IDs (replacing previous file- prefix support)file- prefixfileIdPrefixes is undefined, gracefully falling back to base64 processing87cf954: feat(provider/openai): add support for prompt_cache_key
a3d98a9: feat(provider/openai): add support for safety_identifier
110d167: fix(openai): add missing file_search_call handlers in responses streaming
8d3c747: chore(openai): remove deprecated GPT-4.5-preview models and improve autocomplete control
Updated dependencies [38ac190]
a274b01: refactor(provider/openai): restructure files
Updated dependencies [ a9cea74 ]
### Patch Changes - Updated dependencies [8533108] - Updated dependencies [31205a4] - @ai-sdk/anthropic@3.0.111 - @ai-sdk/provider-utils@4.0.46
minimal in reasoningEffort for openai chatUpdated dependencies [ b74971f ]
Updated dependencies [ 069a945 ]
### Patch Changes - Updated dependencies [31205a4] - @ai-sdk/provider-utils@4.0.46
c9e0f52: Files from the OpenAI Files API are now supported, mirroring functionality of OpenAI Chat and Responses API, respectively. Also, the AI SDK s
Updated dependencies [ f3c1980 ]
@ai-sdk/openai-compatible@2.0.69
@ai-sdk/openai-compatible@2.0.68
5604dde : feat(zai): add the Z.AI provider with GLM chat completions, streaming, reasoning, tools, and multimodal inputs
d5f588f: AI SDK 5
cc62234: chore (provider/openai): switch default to openai responses api
516be5b: ### Move Image Model Settings into generate options
Image Models no longer have settings. Instead, maxImagesPerCall can be passed directly to generateImage(). All other image settings can be passed to providerOptions[provider].
Before
await generateImage({
model: luma.image("photon-flash-1", {
maxImagesPerCall: 5,
pollIntervalMillis: 500,
}),
prompt,
n: 10,
});
After
await generateImage({
model: luma.image("photon-flash-1"),
prompt,
n: 10,
maxImagesPerCall: 5,
providerOptions: {
luma: { pollIntervalMillis: 5 },
},
});
Pull Request: https://github.com/vercel/ai/pull/6180
efc3a62: fix (provider/openai): default strict mode to false
948b755: chore(providers/openai): convert to providerOptions
d63bcbc: feat (provider/openai): o4 updates for responses api
3bd3c0b: chore(providers/openai): update embedding model to use providerOptions
5d959e7: refactor: updated openai + anthropic tool use server side
0eee6a8: Fix streaming and reconstruction of reasoning summary parts
177526b: chore(providers/openai-transcription): switch to providerOptions
2f542fa: Add reasoning-part-finish parts for reasoning models in the responses API
c15dfbf: feat (providers/openai): add gpt-image-1 model id to image settings
3b1ea10: adding support for gpt-4o-search-preview and handling unsupported parameters
e2aceaf: feat: add raw chunk support
d2af019: feat (providers/openai): add gpt-4.1 models
eb173f1: chore (providers): remove model shorthand deprecation warnings
209256d: Add missing file_search tool support to OpenAI Responses API
faea29f: fix (provider/openai): multi-step reasoning with text
7032dc5: feat(openai): add priority processing service tier support
870c5c0: feat (providers/openai): add o3 and o4-mini models
db72adc: chore(providers/openai): update completion model to use providerOptions
a166433: feat: add transcription with experimental_transcribe
26735b5: chore(embedding-model): add v2 interface
443d8ec: feat(embedding-model-v2): add response body field
8d12da5: feat(provider/openai): add serviceTier option for flex processing
9bf7291: chore(providers/openai): enable structuredOutputs by default & switch to provider option
d521cda: feat(openai): add file_search filters and update field names
66962ed: fix(packages): export node10 compatible types
442be08: fix: propagate openai transcription fixes
0059ee2: fix(openai): update file_search fields to match API changes
8493141: feat (providers/openai): add support for reasoning summaries
9301f86: refactor (image-model): rename ImageModelV1 to ImageModelV2
0a87932: core (ai): change transcription model mimeType to mediaType
8aa9e20: feat: add speech with experimental_generateSpeech
4617fab: chore(embedding-models): remove remaining settings
b5a0e32: fix (provider/openai): correct default for chat model strict mode
136819b: chore(providers/openai): re-introduce logprobs as providerMetadata
52ce942: chore(providers/openai): remove & enable strict compatibility by default
db64cbe: fix (provider/openai): multi-step reasoning with tool calls
b3c3450: feat (provider/openai): add support for encrypted_reasoning to responses api
48249c4: Do not warn if empty text is the first part of a reasoning sequence
c7d3b2e: fix (provider/openai): push first reasoning chunk in output item added event
ad2a3d5: feat(provider/openai): add missing reasoning models to responses API
9943464: feat(openai): add file_search_call.results support to include parameter
0fa7414: chore (provider/openai): standardize on itemId in provider metadata
9bd5ab5: feat (provider): add providerMetadata to ImageModelV2 interface (#5977)
The experimental_generateImage method from the ai package now returnes revised prompts for OpenAI's image models.
const prompt = "Santa Claus driving a Cadillac";
const { providerMetadata } = await experimental_generateImage({
model: openai.image("dall-e-3"),
prompt,
});
const revisedPrompt = providerMetadata.openai.images[0]?.revisedPrompt;
console.log({
prompt,
revisedPrompt,
});
fa758ea: feat(provider/openai): add o3 & o4-mini with developer systemMessageMode
d1a034f: feature: using Zod 4 for internal stuff
fd65bc6: chore(embedding-model-v2): rename rawResponse to response
e497698: fix (provider/openai): handle responses api errors
928fadf: fix(providers/openai): logprobs for stream alongside completion model
0a87932: fix (provider/openai): increase transcription model resilience
5147e6e: chore(openai): remove simulateStreaming
06bac05: fix (openai): structure output for responses model
205077b: fix: improve Zod compatibility
c2b92cc: chore(openai): remove legacy function calling
284353f: fix(providers/openai): zod parse error with function
6f231db: fix(providers): always use optional instead of mix of nullish for providerOptions
f10304b: feat(tool-calling): don't require the user to have to pass parameters
4af5233: Fix PDF file parts when passed as a string url or Uint8Array
7df7a25: feat (providers/openai): support gpt-image-1 image generation
Updated dependencies
@ai-sdk/provider-utils@3.0.0-canary.19
@ai-sdk/provider-utils@3.0.0-canary.18
516be5b: ### Move Image Model Settings into generate options
516be5b: ### Move Image Model Settings into generate options
Image Models no longer have settings. Instead, maxImagesPerCall can be passed directly to generateImage(). All other image settings can be passed to providerOptions[provider].
Before
await generateImage({
model: luma.image("photon-flash-1", {
maxImagesPerCall: 5,
pollIntervalMillis: 500,
}),
prompt,
n: 10,
});
After
await generateImage({
model: luma.image("photon-flash-1"),
prompt,
n: 10,
maxImagesPerCall: 5,
providerOptions: {
luma: { pollIntervalMillis: 5 },
},
});
Pull Request: https://github.com/vercel/ai/pull/6180
52ce942: chore(providers/openai): remove & enable strict compatibility by default
928fadf: fix(providers/openai): logprobs for stream alongside completion model
136819b: chore(providers/openai): re-introduce logprobs as providerMetadata
136819b: chore(providers/openai): re-introduce logprobs as providerMetadata
9bd5ab5: feat (provider): add providerMetadata to ImageModelV2 interface (#5977)
The experimental_generateImage method from the ai package now returnes revised prompts for OpenAI's image models.
const prompt = "Santa Claus driving a Cadillac";
const { providerMetadata } = await experimental_generateImage({
model: openai.image("dall-e-3"),
prompt,
});
const revisedPrompt = providerMetadata.openai.images[0]?.revisedPrompt;
console.log({
prompt,
revisedPrompt,
});
284353f: fix(providers/openai): zod parse error with function
Updated dependencies
fa758ea: feat(provider/openai): add o3 & o4-mini with developer systemMessageMode
177526b: chore(providers/openai-transcription): switch to providerOptions
db72adc: chore(providers/openai): update completion model to use providerOptions
ImageModelV1 to ImageModelV28493141: feat (providers/openai): add support for reasoning summaries
3bd3c0b: chore(providers/openai): update embedding model to use providerOptions
d63bcbc: feat (provider/openai): o4 updates for responses api
8aa9e20: feat: add speech with experimental_generateSpeech
26735b5: chore(embedding-model): add v2 interface
948b755: chore(providers/openai): convert to providerOptions
@ai-sdk/provider@2.0.0-canary.4
@ai-sdk/provider@2.0.0-canary.3
a166433: feat: add transcription with experimental_transcribe
Your coding agent can read these notes before it upgrades. Set up the MCP server →