---
name: Changelog
slug: groq-changelog
type: scrape
source_url: https://console.groq.com/docs/changelog
organization: Groq
organization_slug: groq
total_releases: 33
latest_date: 2026-04-18
last_updated: 2026-08-29
tracking_since: 2025-04-14
canonical: https://releases.sh/groq/groq-changelog
organization_url: https://releases.sh/groq
---

<Release date="April 18, 2026" published="2026-04-18T00:00:00.000Z" url="https://console.groq.com/docs/changelog#python-sdk-v120-and-typescript-sdk-v112">
## Python SDK v1.2.0 and TypeScript SDK v1.1.2

Following the [v1.0.0 GA in December 2025](https://github.com/groq/groq-python/releases/tag/v1.0.0), both SDKs received a series of updates over Q1.

**Python SDK** ([v1.2.0](https://github.com/groq/groq-python/releases/tag/v1.2.0))

* **v1.2.0** — Preserve hardcoded query params when merging with user params; ensure file data is sent as a single parameter; multipart request file-copy performance improvements; indices array format for query and form serialization.
* **v1.1.2** — Sanitize endpoint path params; do not pass `by_alias` to Pydantic unless explicitly set; bumped minimum `typing-extensions`.
* **v1.1.0** — Added support for binary request streaming and a custom JSON encoder for extended type support; deprecated Python 3.9.

**TypeScript SDK** ([v1.1.2](https://github.com/groq/groq-typescript/releases/tag/v1.1.2))

* **v1.1.1** — Restored streaming support in `defaultParseResponse`; fixed an abort-signal memory leak and avoided removing abort listeners too early; preserve URL params already embedded in path; pinned patched `minimatch` versions to address CVE-2026-27….
* **v1.1.2** — GitHub Actions OIDC token migration; dependency updates.
</Release>

<Release date="April 18, 2026" published="2026-04-18T00:00:00.000Z" url="https://console.groq.com/docs/changelog#new-voices-for-orpheus-arabic-saudi">
## New Voices for Orpheus Arabic Saudi

Two new voices are available for [canopylabs/orpheus-arabic-saudi](https://console.groq.com/docs/model/canopylabs/orpheus-arabic-saudi):

* **Abdullah** — now the default voice
* **Aisha**

The Arabic Saudi model now supports six voices total: Abdullah, Fahad, Sultan, Lulwa, Noura, and Aisha. Audio samples and example usage are available on the [Orpheus TTS docs](https://console.groq.com/docs/text-to-speech/orpheus).
</Release>

<Release date="April 18, 2026" published="2026-04-18T00:00:00.000Z" url="https://console.groq.com/docs/changelog#minimax-m25-and-qwen3vl-32b-instruct-enterprise">
## MiniMax M2.5 and Qwen3-VL 32B Instruct (Enterprise)

**Added**
- `minimaxai/minimax-m2.5` and `qwen/qwen3-vl-32b-instruct` now available on GroqCloud for Enterprise customers. MiniMax M2.5 is a general-purpose model; Qwen3-VL 32B Instruct is a vision-language model for multimodal use cases. Contact your Groq account team for access.
- New voices for `canopylabs/orpheus-arabic-saudi`: **Abdullah** (now default) and **Aisha**, joining Fahad, Sultan, Lulwa, and Noura for six voices total.

**Changed**
- Python SDK v1.2.0: preserve hardcoded query params when merging with user params; ensure file data sent as single parameter; multipart request file-copy performance improvements; indices array format for query/form serialization. (Also v1.1.2: sanitized endpoint path params, avoided passing `by_alias` to Pydantic unless set, bumped min `typing-extensions`. v1.1.0: added binary request streaming support and custom JSON encoder, deprecated Python 3.9.)
- TypeScript SDK v1.1.2: restored streaming support in `defaultParseResponse`, fixed abort-signal memory leak, preserved URL params already embedded in path, pinned patched `minimatch` versions (CVE-2026-27...); GitHub Actions OIDC token migration and dependency updates.
</Release>

<Release date="January 30, 2026" published="2026-01-30T00:00:00.000Z" url="https://console.groq.com/docs/changelog#platformwide-migration-from-playai-to-orpheus-tts">
## Platform-wide Migration from PlayAI to Orpheus TTS

Following the December 2025 deprecation announcement for `playai-tts` and `playai-tts-arabic`, text-to-speech has been migrated platform-wide to the Orpheus models from Canopy Labs, offering enhanced expressiveness with vocal direction controls, faster inference, and improved audio quality.

Groq Chat now uses Orpheus, with voices:
- English (`canopylabs/orpheus-v1-english`): autumn, diana, hannah, austin, daniel, troy
- Arabic (`canopylabs/orpheus-arabic-saudi`): fahad, sultan, lulwa, noura

Users still on `playai-tts` or `playai-tts-arabic` should migrate before the shutdown date listed on the deprecations page.
</Release>

<Release date="December 1, 2025" published="2025-12-01T00:00:00.000Z" url="https://console.groq.com/docs/changelog#mcp-connectors-beta">
## MCP Connectors (Beta)

MCP Connectors provide a streamlined way to integrate with popular business applications without building custom MCP servers. Groq now supports Google Workspace connectors for Gmail, Google Calendar, and Google Drive via pre-built MCP integrations.

**Available Connectors:** Gmail (read/search emails), Google Calendar (view events), Google Drive (search/access files).

**Key Features:** zero configuration, OAuth 2.0 authentication, OpenAI Responses API compatibility.

**Available Tools:** Gmail (`get_profile`, `search_emails`, `get_recent_emails`, `read_email`); Calendar (`get_profile`, `search`, `search_events`, `read_event`); Drive (`get_profile`, `search`, `recent_documents`, `fetch`).

Currently in beta; feedback welcome via the Community.
</Release>

<Release date="October 29, 2025" published="2025-10-29T00:00:00.000Z" url="https://console.groq.com/docs/changelog#openai-gptosssafeguard-20b">
## OpenAI GPT-OSS-Safeguard 20B

GPT-OSS-Safeguard 20B is OpenAI's first open weight reasoning model trained specifically for safety classification tasks, fine-tuned from GPT-OSS, enabling bring-your-own-policy Trust & Safety AI.

**Key Features:** 131K context window; 65K max output tokens; ~1000 TPS; prompt caching enabled (50% cost savings, $0.037/M vs $0.075/M); Harmony response format with low/medium/high reasoning effort; supports tool use, browser search, code execution, JSON modes, content moderation.

**Use Cases:** Trust & Safety content moderation, policy-based classification, automated triage, policy testing.

**Best Practices:** structure policies with Instructions/Definitions/Criteria/Examples sections, keep policies 400-600 tokens, place static content first for caching, use low effort for simple classifications.
</Release>

<Release date="October 21, 2025" published="2025-10-21T00:00:00.000Z" url="https://console.groq.com/docs/changelog#python-sdk-v0330-typescript-sdk-v0340">
## Python SDK v0.33.0, TypeScript SDK v0.34.0

The Python SDK has been updated to v0.33.0 and the TypeScript SDK has been updated to v0.34.0.

**Key Changes:**

* Improved prompt caching support
* Added annotation/citation support to chat completion messages and streamed deltas
</Release>

<Release date="October 21, 2025" published="2025-10-21T00:00:00.000Z" url="https://console.groq.com/docs/changelog#prompt-caching-enabled-for-gptoss-120b">
## Prompt Caching Enabled for GPT-OSS 120B

**Added**
Automatic prompt caching now live for `openai/gpt-oss-120b`: 50% cost savings on cached input tokens ($0.075/M vs $0.15/M), lower latency, and higher effective rate limits since cached tokens don't count toward limits. Zero setup required.

**Changed**
Python SDK updated to v0.33.0, TypeScript SDK to v0.34.0 — improved prompt caching support and added annotation/citation support to chat completion messages and streamed deltas.
</Release>

<Release date="September 25, 2025" published="2025-09-25T00:00:00.000Z" url="https://console.groq.com/docs/changelog#prompt-caching-enabled-for-gptoss-20b">
## Prompt Caching Enabled for GPT-OSS 20B

Automatic prompt caching now live for `openai/gpt-oss-20b`: 50% cost savings on cached input tokens ($0.037/M vs $0.075/M), lower latency, and automatic prefix matching. Zero setup required.
</Release>

<Release date="September 23, 2025" published="2025-09-23T00:00:00.000Z" url="https://console.groq.com/docs/changelog#remote-model-context-protocol-mcp">
## Remote Model Context Protocol (MCP)

Remote MCP server integration now available in Beta on GroqCloud, connecting AI models to thousands of external tools via Anthropic's open MCP standard. Fully compatible with OpenAI Responses API and OpenAI remote MCP spec, enabling zero-code-change migration from OpenAI to Groq.

**Why it matters:** universal interface to remote MCP servers, faster execution, lower costs, seamless migration.

**Supported Models:** openai/gpt-oss-20b, openai/gpt-oss-120b, moonshotai/kimi-k2-instruct-0905, qwen/qwen3-32b, meta-llama/llama-4-maverick-17b-128e-instruct, meta-llama/llama-4-scout-17b-16e-instruct, llama-3.3-70b-versatile, llama-3.1-8b-instant.

Launch partner tutorials available for BrowserBase, Browser Use, Exa, Firecrawl, HuggingFace, Parallel, Stripe, and Tavily MCP servers.
</Release>

<Release date="September 5, 2025" published="2025-09-05T00:00:00.000Z" url="https://console.groq.com/docs/changelog#moonshot-ai-kimi-k2-instruct-0905">
## Moonshot AI Kimi K2 Instruct 0905

Kimi K2-0905 brings Moonshot AI's model to GroqCloud with day zero support, delivering production-grade speed, low latency and predictable cost for agentic coding.

**Key Features:** 256K context window (largest on GroqCloud to date); prompt caching (up to 50% savings, faster responses); leading price-to-performance at 200+ t/s for $1.50/M tokens blended ($1.00/M input; $3.00/M output); improved agentic coding reliability especially in multi-turn interactions.
</Release>

<Release date="September 4, 2025" published="2025-09-04T00:00:00.000Z" url="https://console.groq.com/docs/changelog#python-sdk-v0311-typescript-sdk-v0320">
## Python SDK v0.31.1, TypeScript SDK v0.32.0

The Python SDK has been updated to v0.31.1 and the Typescript SDK has been updated to v0.32.0.

**Key Changes:**

* Improved chat completion message type definitions for better compatibility with OpenAI. This fixes errors in certain cases with different message formats.
* Added support for new types of Groq Compound tools (Wolfram Alpha, Browser Automation, Visit Website)
</Release>

<Release date="September 4, 2025" published="2025-09-04T00:00:00.000Z" url="https://console.groq.com/docs/changelog#groq-compound-and-compound-mini">
## Groq Compound and Compound Mini

**Added**
Compound (`groq/compound`) and Compound Mini (`groq/compound-mini`) move from beta to general availability as production-ready agentic AI systems integrating web search, code execution, and browser automation in a single API call. Built on GPT-OSS-120B and Llama models, delivering ~25% higher accuracy and ~50% fewer mistakes than OpenAI's Web Search Preview and Perplexity Sonar.

**Key Features:** built-in server-side tools (web search, code execution, Wolfram Alpha, parallel browser automation up to 10 browsers); production-grade stability with increased rate limits; frontier performance on SimpleQA/RealtimeEval; single API call orchestration.

**Changed**
Python SDK v0.31.1, TypeScript SDK v0.32.0 — improved chat completion message type definitions for OpenAI compatibility; added support for new Compound tool types (Wolfram Alpha, Browser Automation, Visit Website).
</Release>

<Release date="August 20, 2025" published="2025-08-20T00:00:00.000Z" url="https://console.groq.com/docs/changelog#prompt-caching">
## Prompt Caching

Prompt caching automatically reuses computation from recent requests sharing a common prefix, delivering cost savings and improved response times with volatile-only storage that expires automatically.

**How It Works:** prefix matching against recently processed requests; cache hit reuses computation reducing latency and cutting token costs by 50% for cached portions; cache miss processes normally and caches the prefix; automatic expiration within a few hours.

Rolling out first to Kimi K2, with more models coming. No code changes or additional fees required.
</Release>

<Release date="August 5, 2025" published="2025-08-05T00:00:00.000Z" url="https://console.groq.com/docs/changelog#python-sdk-v0310-typescript-sdk-v0300">
## Python SDK v0.31.0, TypeScript SDK v0.30.0

The Python SDK has been updated to v0.30.0 and the Typescript SDK has been updated to v0.27.0.

**Key Changes:**

* Added support for `high`, `medium`, and `low` options for `reasoning_effort` when using GPT-OSS models to control their reasoning output.
* Added support for `browser_search` and `code_interpreter` as function/tool definition types in the `tools` array in a chat completion request. Specify one or both of these as tools to allow GPT-OSS models to automatically call them on the server side when needed.
* Added an optional `include_reasoning` boolean option to chat completion requests to allow configuring if the model returns a response in a `reasoning` field or not.
</Release>

<Release date="August 5, 2025" published="2025-08-05T00:00:00.000Z" url="https://console.groq.com/docs/changelog#responses-api-beta">
## Responses API (Beta)

[Groq's Responses API](https://console.groq.com/docs/responses-api) is fully compatible with OpenAI's Responses API, making it easy to integrate advanced conversational AI capabilities into your applications. The Responses API supports both text and image inputs while producing text outputs, stateful conversations, and function calling to connect with external systems.

This feature is in beta right now — please let us know your feedback on our [Community Forum](https://community.groq.com/)!
</Release>

<Release date="August 5, 2025" published="2025-08-05T00:00:00.000Z" url="https://console.groq.com/docs/changelog#openai-gptoss-20b--openai-gptoss-120b">
## OpenAI GPT-OSS 20B & OpenAI GPT-OSS 120B

**Added**
GPT-OSS 20B and GPT-OSS 120B are OpenAI's open-source Mixture-of-Experts models performing on par with o4-mini/o3-mini, with reasoning, built-in browser search and code execution, and structured outputs support.

**Key Features:** 131K context window; 32K max output tokens; ~1000+/~500+ TPS; MoE with 32/128 experts respectively; surpasses o4-mini on many benchmarks.

**Performance (20B):** 85.3% MMLU, 60.7% SWE-Bench Verified, 98.7% AIME 2025, 75.7% avg MMMLU.
**Performance (120B):** 90.0% MMLU, 62.4% SWE-Bench Verified, 57.6% HealthBench Realistic, 81.3% avg MMMLU.

Also added: **Responses API (Beta)** — fully compatible with OpenAI's Responses API, supporting text/image inputs, stateful conversations, and function calling.

**Changed**
Python SDK v0.30.0/v0.31.0, TypeScript SDK v0.27.0 — added `reasoning_effort` (high/medium/low) for GPT-OSS models; added `browser_search` and `code_interpreter` tool types; added optional `include_reasoning` boolean for chat completions.
</Release>

<Release date="July 18, 2025" published="2025-07-18T00:00:00.000Z" url="https://console.groq.com/docs/changelog#structured-outputs">
## Structured Outputs

Groq now supports structured outputs with JSON schema for `moonshotai/kimi-k2-instruct`, `meta-llama/llama-4-maverick-17b-128e-instruct`, and `meta-llama/llama-4-scout-17b-16e-instruct`. Guarantees model responses strictly conform to a provided JSON Schema, ensuring reliable data structures without missing fields or invalid values, eliminating complex parsing logic.

**Key Benefits:** guaranteed compliance, type safety, improved developer experience without prompt engineering for format adherence.
</Release>

<Release date="July 15, 2025" published="2025-07-15T00:00:00.000Z" url="https://console.groq.com/docs/changelog#moonshot-ai-kimi-2-instruct">
## Moonshot AI Kimi 2 Instruct

[Kimi K2 Instruct](https://console.groq.com/docs/model/moonshotai/kimi-k2-instruct) is Moonshot AI's state-of-the-art Mixture-of-Experts (MoE) language model with 1 trillion total parameters and 32 billion activated parameters. Designed for agentic intelligence, it excels at tool use, coding, and autonomous problem-solving across diverse domains.

**Key Features:**

* 131K token context window
* 16K max output tokens
* MoE architecture with 384 experts (8 selected per token)
* Surpasses GPT-4.1 on agentic and coding use cases

**Performance Metrics:**

* 53.7% Pass@1 on LiveCodeBench (coding performance)
* 65.8% single-attempt accuracy on SWE-bench Verified
* 89.5% exact match on MMLU
* 70.6% Avg@4 on Tau2 retail tasks
</Release>

<Release date="July 15, 2025" published="2025-07-15T00:00:00.000Z" url="https://console.groq.com/docs/changelog#python-sdk-v0300-typescript-sdk-v0270">
## Python SDK v0.30.0, TypeScript SDK v0.27.0

**Changed**
Python SDK updated to v0.30.0, TypeScript SDK to v0.27.0 — improved chat completion message type definitions for better OpenAI compatibility, fixing errors with certain message formats.

**Added**
Moonshot AI Kimi K2 Instruct: a state-of-the-art Mixture-of-Experts model with 1 trillion total parameters and 32 billion activated parameters, designed for agentic intelligence — tool use, coding, autonomous problem-solving.

**Key Features:** 131K context window; 16K max output tokens; MoE with 384 experts (8 selected per token); surpasses GPT-4.1 on agentic and coding use cases.

**Performance:** 53.7% Pass@1 LiveCodeBench, 65.8% SWE-bench Verified, 89.5% MMLU exact match, 70.6% Avg@4 Tau2 retail tasks.
</Release>

<Pagination cursor="2025-07-15T00:00:00.000Z|2026-08-11T19:28:35.442Z|rel_3iUxNVViu8-VrqIOqSTxP" next="https://releases.sh/groq/groq-changelog.md?cursor=2025-07-15T00%3A00%3A00.000Z%7C2026-08-11T19%3A28%3A35.442Z%7Crel_3iUxNVViu8-VrqIOqSTxP&limit=20" />