Release of Mistral Small 4
Mistral Small 4 (mistral-small-2603) is a hybrid model that unifies instruct, reasoning, and coding capabilities in a single multimodal model with a 256k context window.
Also shipped Mar 16, 2026Mistral AI in March 2026 โ
๐ค New LLM Models Support & Bug Fixes
We've added support for new LLM models in the Voice Agent API: OpenAI GPT-5.3 Instant ( gpt-5.3-chat-latest ) OpenAI GPT 5.4 ( gpt-5.4 ) Google Gemini 3.1 Flash Lite ( gemini-3.1-flash-lite-preview ) Example: { "type": "Settings", "agent": { "think": { "provider": { "type": "open_ai", "model": "gpt-5.3-chat-latest" } } } } For the full list of supported...
Metrics Open Beta Now Supports Alerting
Metrics, currently in open beta, now support creating alerts.
Persisting Table Filter in URL & Persisting Collapsed Card State in Experiment/Flags UI
What it is: Two state persistence improvements for Experiment and Feature Flags. Experiment table filters now encode into the URL, so filtering by status, tag, or owner produces a link you can bookmark or send to a teammate.
Targeting based on First Seen Timestamp
What it is: A built-in "First Seen" user property is now available in Guides and Surveys' remote evaluation. Cohort-based new user targeting has a reliability problem: membership updates run with a delay, so a brand-new user won't be included until after their first session has already passed.
LlamaIndex 0.14.18
Release Notes [2026-03-16] llama-index-agent-agentmesh [0.2.0] chore: deprecate python 3.9 once and for all ( #20956 ) llama-index-agent-azure [0.3.0] chore: deprecate python 3.9 once and for all ( #20956 ) llama-index-callbacks-agentops [0.5.0] chore: deprecate python 3.9 once and for all ( #20956 ) llama-index-callbacks-argilla [0.5.0] chore: deprecate...
Guardrails 0.9.2
- Bump lodash-es from 4.17.21 to 4.17.23 in /docs in the npm_and_yarn group across 1 directory by @dependabot [bot] in #1411 Bump pypdf from 6.7.0 to 6.7.1 in the pip group across 1 directory by @dependabot [bot] in #1413 Fix ModuleNotFoundError when installing hub validators dynamically [feat]: DB Commands fix: preserve falsy values in safe_get [chore]...
More from Mistral AI Studio and APIEvery Mistral AI Studio and API release โ
Release of Leanstral Code Agent
Leanstral (labs-leanstral-2603) is an open-source code agent designed for Lean 4 formal proof engineering.
Release of Mistral Moderation 2603
Mistral Moderation 2603 (mistral-moderation-2603) has been released.
Custom Guardrails Support for Agents and Conversations
Custom Guardrails support has been added for Agents and Conversations. Guardrails can be configured directly on an Agent via the guardrails parameter, or passed per-request on conversations and chat completions endpoints.
Release of Voxtral TTS
Voxtral TTS (voxtral-tts-2603) is a text-to-speech model featuring zero-shot voice cloning, multilingual support, and real-time streaming.
Release of Voxtral Mini Transcribe 2 and Realtime
Voxtral Mini Transcribe 2 (voxtral-mini-2602) and Voxtral Mini Transcribe Realtime (voxtral-mini-transcribe-realtime-2602) have been released.
Context Biasing in Audio Transcriptions API
Context biasing has been introduced in the Audio Transcriptions API.
Diarization in Audio Transcriptions API
Diarization support has been introduced in the Audio Transcriptions API.
Release of Mistral Medium 3.5
Mistral Medium 3.5 (mistral-medium-3-5) is a frontier-class multimodal model optimized for agentic and coding use cases. It features adjustable reasoning via the reasoning_effort parameter and is released under a Modified MIT license.