Updated monthly

Changelog

New models, features, and API changes shipped to Venice each month.

Changelog

March 2025

Models

12 changes

DeepSeek Coder V2 Lite
Released DeepSeek Coder V2 Lite as a Code model for all users.
All users
Mistral Small 3.1 24B
Mistral Small 3.1 24B is now available with a 128k token context limit. It supports vision, function calling, and web search, and is one of the fastest large-context models on Venice.
Pro users
Qwen QwQ 32B
Qwen QwQ 32B is now available for Pro and API users. It is a reasoning model capable of thinking and reasoning, offering competitive performance against models like DeepSeek-R1 and o1-mini. The implementation supports web search and structured responses.
Pro users
Deepseek Coder V2 Lite
Re-activated the Deepseek Coder V2 Lite beta.
Beta
In-Painting Vision Model
Updated In-Painting to use Mistral as the vision model for mask generation, improving performance of in-painting requests.
All users
Qwen and Mistral Censorship
Updated Venice system prompts to address censorship issues appearing in Qwen and Mistral models.
All users

50 changes

Character Voice Support
Character creators can now assign a voice and speed to their characters. Pro users with Venice Voice enabled will hear character responses read aloud with the selected voice.
Pro users
Character web toggle
Character chat now includes a web-enabled toggle as part of the input.
High Temperature Notice
Added a notice when temperature is greater than 1, warning that high temperature may create gibberish responses.
Jump to Latest Message
Added a feature to jump directly to the latest message in long conversations. This feature received 52 votes on Featurebase.
LLM-Generated Link Warning
Added a warning when users click on links generated by LLMs.
Multiple Document Support
Launched support for uploading multiple documents in a single chat. This feature was requested by users and received 162 votes on Featurebase.

EXIF Metadata in Generated Images
Generated images now include EXIF metadata with generation settings. This can be disabled in Advanced Image Settings.
Image Carousel View
Added a carousel view for viewing multiple generated images in a single batch.
Multi-Image Variants
Pro users can now generate 1 to 4 image variants in a single request, configurable via the Image Settings panel. Clicking a variant shows a zoomed view with action buttons available.
Pro users
Venice Voice Processes 1M Sentences
Venice Voice processed over 1 million sentences in its first 24 hours. API access to voice capabilities will be available in the coming weeks.
Voice Download
Users can now download audio generated by Venice Voice as a .wav file once generation is complete, using a download icon that appears next to the speaker.
EXIF Includes Model Name
Generated images now include the model name in the EXIF data, providing better traceability of image origin.

Character Image Safe Overlay
Fixed an issue where blurred character images generated with the character generator were not showing the Safe Venice description overlay.

Safe Venice mode
When the account level “Safe Venice” mode is enabled, uncensored image models are redacted from the model selector.
VCU Chart Default
Adjusted the VCU chart to default to a 90-day view for better visibility of usage trends.

Consumption Limits for API Keys
Users can now define consumption limits per epoch in both USD and VCU on API keys via the web UI or API, enabling fine-grained allocation of resources when sharing VCU with others.
Sticker Factory Price Reduction
Venice Sticker Factory prices have been reduced to $9.99.

Venice Brand Page
Launched a new brand page at https://venice.ai/brand with official assets and guidelines.
Venice Rebrand
Venice has been rebranded, and 1/3 of the VVV token supply was burned as part of the update.
App Loading Screen
Rewrote the app loading screen to eliminate flashes and glitches during initial load, introducing a smooth fade transition.
EXIF Data on Upscaled Images
Upscaled images now retain their EXIF data. This helps preserve metadata through the upscaling process.
Next.js Upgrade
Upgraded the app to Next.js version 15, improving performance and reliability of the app platform.
Rate Limit Debit Logic
Adjusted rate limits so that users are only debited for successful API requests.

Mobile Wallet Sign-In
Mobile wallets like Coinbase Wallet and Metamask are now redirected to the sign-in flow when visiting the Venice homepage, reducing login friction.

38 changes

Optimized for Code Flag
The model list endpoint now exposes an optimizedForCode capability on text models.
Supports Vision Flag
The model list endpoint now exposes a supportsVision capability on text models.
supportsReasoning Capability
Reasoning models now expose a supportsReasoning field in the models list endpoint to indicate reasoning capability.
Longer Prompts on Flux
Increased the length of supported prompts on flux-uncensored models via the API.
EXIF Metadata Control
Added the embed_exif_metadata parameter to control whether EXIF metadata is included in generated images via API.
Image Constraints Exposure
The model list endpoint now exposes all image constraints that are part of the schema validation for image generation.

23 changes

Character Context Bug
Fixed a bug where returning to a character after an image conversation caused an image generation in the first message with that character.
Chat
Character Info UI Issues
Fixed several UI display issues on mobile for character info and initial character display pages.
Mobile
Markdown Rendering Issues
Fixed issues with Markdown rendering in chat, particularly with ordered lists and code blocks.
Chat
Markdown Upload
Fixed a bug preventing Markdown files from being uploaded to the Text inference endpoint.
Chat
Ordered List Rendering
Fixed a bug where ordered lists generated by LLMs might render incorrectly.
Chat
Pro Status Errors
Fixed a bug where Pro users could see erroneous messages about inference requests requiring Pro status under certain conditions.
Pro usersChat

Share feature requests and vote on ideas from the community on Learn Venice.