New Models

Text Models

• Claude Fable 5.1 — Text model from Anthropic. Anonymous. Available to all users

• GLM 5.3 Flash — Frontier text model from Zhipu AI. Private. Available to all users

• GLM 5.3 — Frontier text model from Zhipu AI. Private. Available to all users

• Qwen 3.8 27B — 27-billion-parameter text model from Alibaba Cloud in the Qwen 3.8 family. Private. Available to all users

• Gemini 3.7 Flash — Latency-optimized text model from Google DeepMind. Anonymous. Available to all users

• DeepSeek V4 Pro 0813 — Text model from DeepSeek, August 13 release of the V4 Pro series. Private. Available to all users

• Qwen 3.8 2.4T — 2.4-trillion-parameter text model from Alibaba Cloud in the Qwen 3.8 family. Private. Available to all users

• Grok 4.6 — Text model. Private. Available to all users

• DeepSeek V4 Flash 0731 Fast — Speed-optimized variant of DeepSeek's V4 Flash (July 31 release). Private. Available to all users

• Kimi K3 Fast — Speed-optimized text model from Moonshot AI in the K3 family. Private. Available to all users

• DeepSeek V4 Flash 0731 — Text model from DeepSeek, July 31 release of the V4 Flash series. Private. Available to all users

• Qwen 3.8 Max — Largest-scale variant in Alibaba Cloud's Qwen 3.8 family. Anonymous. Available to all users

Image Models

• Grok Imagine 2.0 — Image generation model with support for photorealistic and stylized outputs. Private.

• Qwen Image 3 — Alibaba Cloud's image generation model with text rendering and multi-style output capabilities. Anonymous.

• Qwen Image 3 Edit — Alibaba Cloud's image editing variant of Qwen Image 3, supporting instruction-based modifications to existing images. Anonymous.

• Qwen Image 3 Pro — Higher-fidelity variant of Alibaba Cloud's Qwen Image 3 with improved detail, coherence, and prompt adherence. Anonymous.

• Qwen Image 3 Pro Edit — Editing variant of Qwen Image 3 Pro from Alibaba Cloud, combining Pro-tier quality with instruction-based image modification. Anonymous.

Video Models

• Gemini Omni Flash 1.1 — Video generation model from Google DeepMind. Anonymous. Available to all users

• Gemini Omni Flash 1.1 R2V — Reference-to-video variant of Gemini Omni Flash 1.1 from Google DeepMind, enabling image-guided video generation. Anonymous. Available to all users

• MiniMax H3 Max R2V — Reference-to-video variant of MiniMax H3 Max, enabling image-guided video generation. Private. Available to all users

• MiniMax H3 Max — Video generation model from MiniMax. Private. Available to all users

• Wan 3.0 Prime — Text-to-video generation model producing high-fidelity video from text prompts. Anonymous. Available to all users

• Wan 3.0 Prime Reference — Reference-to-video variant of Wan 3.0 Prime, enabling image-guided video generation. Anonymous. Available to all users

• Wan 2.2 Enhanced — Enhanced variant of the Wan 2.2 video generation model with improved output quality. Anonymous. Available to all users

• LTX Video 2.5 Fast — Lightricks video generation model optimized for faster generation speed. Anonymous. Available to all users

• LTX Video 2.5 Pro — Lightricks video generation model optimized for higher output quality. Anonymous. Available to all users

• Seedance 2.5 — Text-to-video generation model with high-quality motion and scene coherence. Anonymous. Available to all users

• Seedance 2.5 R2V — Reference-to-video variant of Seedance 2.5, enabling image-guided video generation. Anonymous. Available to all users

• MiniMax H3 Enhanced — Enhanced video generation model from MiniMax with improved visual fidelity. Anonymous. Available to all users

• MiniMax H3 R2V Enhanced — Reference-to-video variant of MiniMax H3 Enhanced from MiniMax, enabling image-guided video generation. Anonymous. Available to all users

• Flux 3 — Video model from Black Forest Labs. Anonymous. Available to all users

• Flux 3 First Last Frame — Black Forest Labs video model that generates video by interpolating between user-provided first and last frames. Anonymous

• Wan 3.0 — Alibaba Cloud video generation model. Anonymous

• Wan 3.0 Reference — Alibaba Cloud video model with reference image conditioning for subject or style consistency in generated video. Anonymous

• MiniMax H3 — Video generation model from MiniMax. Anonymous

• MiniMax H3 R2V — MiniMax video model with reference-to-video capability, generating video conditioned on a reference image. Anonymous

• Grok Imagine 1.5 R2V — xAI video model with reference-to-video generation, producing video guided by a reference image. Private

App

Chat

• Sub-Agent Side Panel — Sub-agent output streams now open in a side panel instead of inline.

• Agent Search — New advanced search interface for finding agents.

• Content Reporting — Users can now report content directly from the interface.

• Reasoning Effort for Kimi K3 — Reasoning effort is now configurable when using the Kimi K3 model.

• LCC Toggle — The LCC toggle now appears on eligible non-pay-per-use models.

• Status Indicator Redesign — The auto-expanding status widget has been replaced with a sidebar indicator that opens a detail modal on click.

• Local PDF Parsing — PDF files are now parsed locally in the browser instead of server-side

• Text Input Character Limits — All text input fields now enforce maximum character length constraints.

• Citation Link Warning — A warning is now shown when clicking external links in citations.

• Sound Effects Prompt Limit — Increased the maximum prompt length for sound effects generation to 450 characters.

• MiniMax H3 Prompt Limit — Increased the prompt character limit for MiniMax H3 to 7,000 characters.

• Qwen 3.8 27B Reasoning Effort — Added reasoning effort control for the Qwen 3.8 27B model.

• Qwen 3.8 2.4T Function Calling — Function calling is now enabled for this model.

• Prompt Enhancer on Edit/Combine — The prompt enhancer now runs automatically when submitting edit or combine operations.

• Chat Composer Overlay — Messages beneath the chat composer now fade out with a gradient mask instead of being hidden behind a solid backdrop.

• Default Thinking Level — The default thinking level for reasoning models is now set to High (previously Normal).

• Chat Memory for TEE Models — Chat Memory is now available on TEE models when end-to-end encryption is turned off

• Publication Rejection Banner — A rejection banner is now displayed when an agent or publication fails moderation review.

• Full-Size File Preview in Chat — Clicking file thumbnails in classic chat now opens a full-size preview.

• Tool Icons in Details Card — Tool icons now appear on the details card for agents and listings.

• Copy JSON in Admin — New "Copy JSON" button on the admin agent detail modal.

Image, Video and Audio

• Video-to-Video Quotes — Video-to-video price quotes now factor in the detected duration of the input video.

• Replace Asset — Added the ability to replace an uploaded asset in place.

• Voice Selection UI — Selected voice and active recording state are now more clearly indicated in the UI.

• Audio Trimming — Audio trimming is now decode-accurate; files that cannot be trimmed are rejected instead of silently failing.

• Preset Voices for Grok Imagine 1.5 R2V — Added preset xAI voice options for the Grok Imagine 1.5 R2V model

• Video Model Annotations — Video model options now display capability annotations matching the format used in the image model selector.

• Prompt Enhancer for Inpainting — The prompt enhancer is now automatically applied when using the inpaint mask tool.

• Audio Studio Loop Toggle — New toggle in Audio Studio to loop audio playback.

• Doodle Mode Updates — Improved model picker in Doodle mode and added a recommendation call-to-action for the Seedream model.

• Studio Reasoning Variants — Generation and edit variants in Studio now support reasoning output.

• Video Price Quotes — Video generation price estimates now factor in the number of reference images included.

• Oversized Image Paste Handling — Pasting an oversized image now automatically resizes it instead of rejecting it.

• New xAI TTS Voices — Additional text-to-speech voices from xAI are now available in Audio Studio.

• Mask-Aware Prompt Enhancer — The prompt enhancer now factors in the inpaint mask region and displays the mask during enhancement.

• Model Selection for Image & Video Tools — New model selector UI added to image and video generation tools.

• Editor Import Metadata — Importing generated media into the editor now carries over the original prompt, model, and settings.

• Reference Audio Limits — Chat and studio now enforce the maximum number of reference audio files per model

• Mobile HTML Sandbox — HTML sandbox previews now render on mobile devices

• Wan 3.0 Reference Image Limit — Upload tiles now enforce the maximum reference image count for Wan 3.0

• Reference Image Deduplication — Duplicate reference images are automatically filtered out when appended

• Video Task ID — Task ID is now displayed in video details panels in both chat and studio

• Reference Image Preview — Click reference image thumbnails to view the full-size image

• Video Card Layout — Video cards now display at full width

• Video Edit UI — Added a new interface for editing generated videos

• Character Photo Generation Settings — Expanded and reworked the configuration options for character photo generation.

• Video Model Ordering — Wan 2.7 Enhanced (Image-to-Video and Text-to-Video) is now pinned to the top of the video model selector.

• Studio Paste Confirmation — The paste confirmation dialog in Studio now clarifies that pasting creates an Edit session.

• Voice Menu Avatars & Previews — The Studio Voice menu now displays seeded gradient avatars and per-voice audio previews.

• Audio Studio Recording — Added source audio input and recording support to Audio Studio.

• Reference Audio Preview — Reference audio clips now support click-to-play inline preview.

• Wan 3.0 Reference-to-Video Audio Input — A reference audio input field is now shown when using Wan 3.0 in Reference-to-Video mode.

• Wan 3.0 Reference Audio in Chat — Classic chat now accepts and auto-truncates reference audio for Wan 3.0 video generation.

• Agentic Video Notifications — Users receive a notification when an agentic video generation completes.

• Model Search in Tool Editor — Added a search field in the tool editor to find models by name.

• Studio Edit History Clustering — Studio edit rail history is now grouped into per-turn clusters instead of a flat list.

• Export Video History — Video generation history can now be exported as CSV or JSON.

• Session Edit in Studio — Sessions can now be edited directly within Studio.

Library and Media

• Uncategorized Category — "Uncategorized" is now a browsable, directly linkable category.

• Publish Confirmation Popover — Publishing now requires confirmation via a stateful review popover showing current status.

• Publishing Progress Indicator — The publish flow now displays progress, and the publish button is disabled when there are no pending changes.

• Listing Search Relocated — The listing search bar has moved from the page body into the header.

• Reference Media Validation — Title and description fields are now required when editing reference media.

Account and Settings

• Discount Expiry Info — Discount tooltips now show the exact expiration date and time.

• Usage Table Layout — The reference ID in the usage table has moved into a per-row overflow menu.

• Progressive Blur Header — Page header now uses a progressive blur overlay effect

• Cost Management Date Ranges — Added 90-day, 6-month, 12-month, year-to-date, and all-time date ranges, plus working custom date windows

• Fair-Use Rate Limit Messaging — Paid users now see a fair-use rate-limit notice when usage thresholds are reached.

• Publishing Rate Limit Toast — A toast notification now appears when a publishing rate limit is reached.

• Creator Social Avatar — Creator profile pages now display the creator's linked social avatar.

Wallet and Payments

• Crypto Monthly Subscriptions — Monthly subscription plans can now be paid via BTCPay or Coinbase

• Crypto Subscriptions General Availability — Monthly crypto-paid subscriptions are now available to all users

• API Image Billing — Image generation via the API now tracks and bills credit usage

• Bitcoin Credit Purchases — Credits can now be purchased using Bitcoin via BTCPay

• USDC Payments via Stripe — USDC is now available as a payment method in the credits purchase modal through Stripe

• Localized Currency Display — Credit purchase amounts now display in the user's local currency format during checkout

Mobile App

• Image View Actions — Added action buttons such as save and share to the image viewer.

• Reference Image Details — Media reference images now display their associated metadata.

• Agentic Chat - Introduced Agentic Chat (Chat V2), a new multi-step chat mode powered by an agent, with access to Venice tools and features.

• Encrypted Chat Backup — Chat history and media can now be backed up and restored with passphrase-based encryption

• Android Native Paste Menu — Android text fields now use the OS-native paste menu

• iOS Native Paste Menu — iOS text fields now use the OS-native paste menu

• Model Section Layout — Text and image model settings are now displayed in separate sections

• iOS Chat Rename — Chat rename prompts on iOS now use native system dialogs

• Account Pill Settings Shortcut — Tapping the account pill now opens Settings directly

• Settings Navigation Overhaul — Redesigned the Settings navigation structure and root screen

• Native Context Menu — Added native OS context menu support for long-press actions

• Per-Model Image Input Limits — Image attachment limits are now enforced per model based on each model's supported input count

• Reference Image Previews — Image previews now display in the reference edit sheet when editing image references.

• Offline Indicator — Added an offline status banner in the header when the device loses connectivity.

• Database Recovery — Added a force-reinitialize option to the locked database dialog when the local database becomes inaccessible.

• Database Backup Prompt — The locked database dialog now prompts users to back up data before taking recovery actions.

• Image Edit Limit Removed — Removed the per-image edit limit; users can now make unlimited edits to a generated image.

• Fork Conversation — Added a fork conversation option in classic chat, allowing users to branch a conversation from any message.

• Glass Theme Colors — Added additional color options to the glass theme.

• Glass Chat Chrome — Added a floating glass-style chat chrome UI for conversations.

• Frosted Glass Chat Header — The chat header now uses a frosted glass visual effect in the mobile app.

• Video Loading Cards — Video generation and video editing outputs now show loading card indicators while processing.

• Voice Chat Ready State — Voice chat now displays an "ask anything" prompt when ready for input.

• Video Edit Accessibility — Improved accessibility support in the video editing interface.

• Agentic Video Edit Results — Agentic video editing now renders edit results inline in the conversation.

• App Status in Header — The app header now surfaces current app status notifications when relevant.

• Chat Carousel Image Swap — Images in the chat carousel can now be swapped out during editing.

• Chat Improvements — User messages now render code blocks; daily-limit messaging for paid users shows local reset times; model list displays a "cheapest" badge.

• Doodle Before Inpaint — Users can now draw on an image before applying inpaint edits.

• Model Explorer from Selector — The model selector now includes an option to open the model explorer.

• Model Info Tap Action — Tapping model info in the selector now opens a model details view.

• xAI TTS Voices — Added xAI text-to-speech models to the voice switcher.

• TTS Model Switcher — New switcher component for selecting between available text-to-speech models.

• Publish Action — Added a Publish button to the media details header

• Rate Limit Banner — A "slow down" banner now appears when the per-minute chat message limit is reached

• Font Options — Added a font selection popover in Settings to choose between available typefaces

• Floating Drawer Header — Navigation drawer now displays a floating header that stays visible while scrolling

• Font Selector — New font picker in settings with added Inter font support

API

• Gemini Omni Flash 1.1 — Gemini Omni Flash 1.1 is now available through the public API.

• Voice Parameter Defaults — Redundant default voice values are no longer returned in API responses.

• Seedance Video API Documentation — Documented source-matched aspect ratio and duration parameters for the Seedance Video API.

• Seedance API Content Policy — Updated media content policy documentation for the Seedance public API.

• API vs App Pricing Docs — Added documentation clarifying the distinction between API pricing and consumer app pricing.

• Billing History API Reference —w Added a new API reference page for usage and billing history.

• Uncensored Model Flag — GET /v1/models now returns model_spec.uncensored, so clients no longer need to infer the label from model traits or model ID substrings.

• Embeddings Text-Only Validation — Embeddings requests that pass token ID arrays now return a validation error instead of failing upstream.

• Embeddings Privacy Enforcement — The PRIVATE_TEXT API key privacy setting now applies to embeddings requests, which were previously treated as non-text.

• Billing Usage Endpoint Sunset — GET /v1/billing/usage is deprecated in favor of GET /v1/billing/usage-history.

• Wan 3.0 General Availability — Wan 3.0 text-to-video, image-to-video, and reference-to-video are now available to all API users,

• Wan 3.0 Prime — Added new Wan 3.0 Prime video model variants to the API.

• Grok Imagine 2.0 — Grok Imagine 2.0 image generation and editing are now available to all API users.

• DeepSeek V4 Flash Fast — DeepSeek V4 Flash 0731 Fast is now available on the API with a maximum output length of 128K tokens.

• Qwen Image 3 Edit Models — Qwen Image 3 Edit and Qwen Image 3 Pro Edit are now available through the image edit API.

• Video Prompt Length Increase — Increased the maximum prompt and negative_prompt length for video requests from 10,000 to 20,000 characters.

• Sound Effects Loop Parameter — The Sound Effects API now accepts a loop parameter.

• TTS Response Format — The speech endpoint now accepts response_format, and GET /v1/models now publishes the output formats supported by each speech model.

• Model Listing Privacy Filter — GET /v1/models now filters results based on the calling API key's model privacy setting and returns MODEL_PRIVACY_RESTRICTED for models the key cannot access.

• Partner Rate Quoting — The cost object in chat completions responses now reflects negotiated partner rates instead of public list rates.

• Model Listing Cache Headers — GET /v1/models responses are no longer cacheable because model visibility and negotiated rates can vary by caller.

• Character API Visibility — Characters can now be individually marked as API-accessible, controlling whether they appear on the characters endpoint.

• Seedance Bitrate Mode — Seedance 2.0 and 2.5 video requests now accept a bitrate_mode of standard or high.

• Rate Limit Error Types — Rate limit errors now include a typed reason, with documented values for RPM, TPM, RPD, FAILED_REQUESTS, and UNSUPPORTED_FEATURE_REQUESTS.

• Paid Tier Text Limits — Adjusted paid tier daily text limits to 10,000 requests per day.

• xAI Video Input Pricing — Video requests to xAI models are now billed per input image instead of using a flat surcharge.

• Image Promotional Pricing — Promotional discounts now apply to image models, including launch discounts for Qwen Image 3 and MiniMax H3.

• Qwen 3.8 Max Price Reduction — Reduced Qwen 3.8 Max list prices by 50% as a limited-time promotion.

• Model Trait Reassignment — Updated model traits to point to active models: default and function_calling_default now resolve to GLM 5.2, most_intelligent to Grok 4.6, default_reasoning to Kimi K3, default_vision to Qwen 3.8 27B, and default_code to DeepSeek V4 Pro 0813.

• Gemini 3.7 Flash — Gemini 3.7 Flash is now available on the API with a 50% introductory discount through December 31, 2026.

• Nemotron Cascade 2 Removed — NVIDIA Nemotron Cascade 2 30B is no longer available through the API.

• Voice Changer Error Handling — Voice Changer requests that fail permanently now return a terminal, non-retryable error instead of a retryable HTTP 500.

• Voice ID Error Handling — Requests with an unresolvable Voice ID now return a terminal error instead of a retryable HTTP 500.

• Models API Promotional Rates — The public models endpoint now includes promotional pricing data when applicable.

• Voice Changer Credit Refunds — Fixed an issue where credits were not refunded after Voice Changer failures due to a truncated charge identifier.

• Overload Retry Headers — Overloaded text-to-image requests now return HTTP 429 with a Retry-After header where supported.

• New TTS Voices — Added new voices to the text-to-speech API.

• Image URL Validation — The API now validates remote image_url values on inference requests and rejects invalid or unreachable URLs.

• Video URL Validation — The API now validates remote video_url values on inference requests and rejects invalid or unreachable URLs.

• Chat Rate Limits for Paid Tiers — Configured chat rate limits are now enforced for paid tier users.

• Per-API-Key Model Privacy Enforcement — Model privacy settings configured per API key are now enforced on inference requests.

• Default API Model — Updated the default model used for API text generation requests.

• API Key Deep Link — Added support for deep-linking directly to the Create API Key modal.

• Wan 3.0 API Defaults — Aligned Wan 3.0 default parameters with the general availability API specification.

• Per-API-Key Model Privacy Setting — API keys now support a persistent model privacy setting.

• Rate Limit Increase — Doubled Pro and Partner tier rate limits for Gemma 4 26B A4B and Qwen3.6 35B A3B.

• Seedance OpenAPI Documentation — Added source-matched duration and aspect ratio parameters to the OpenAPI specification for Seedance video generation.

• Seedance Source-Matched Parameters — The Video API now accepts source-matched aspect ratio and duration parameters for Seedance requests.

• Seedance Edit/Extend Fix — Free-form edit and extend prompts are now detected before request submission, reducing provider rejections caused by mismatched ratio or duration parameters.

• Seedance Faces-Off Endpoints — Added paired API endpoints for Seedance video generation with face detection disabled.

• Model Rate Limit Increases — Increased rate limits by 50% for kimi-k3, deepseek-v4-flash-0731, qwen3-6-27b, and deepseek-v4-flash-0731.

Model Deprecations

• Phala Qwen3.6 35B Uncensored 

• NEAR Gemma 4 31B 

• AionLabs Aion 2.0 

• Arcee Trinity Large (Thinking) 

• GLM 5 (TEE) 

• GLM 4.7 Flash (TEE) 

• Qwen3.5 122B A10B (TEE) 

• LTX-2 and LTX-2 19B 

• Grok Imagine Image Pro 

• Runway Gen-4 Aleph 

• Claude Opus 4.7 Fast 

• GLM 4.7 (TEE)

• Venice Uncensored 24B (TEE) 

• Qwen3 30B A3B (TEE) 

• DeepSeek V4 Flash 

New Models

Text

Image & Video Models

Audio Models

Model Updates

• TEE GLM 5.2 Function Calling — Function calling is now enabled for TEE GLM 5.2 (e2ee-glm-5-2-p)

• Ideogram V4 Rendering Speed — Ideogram V4 is now pinned to QUALITY rendering speed

• Qwen3 6-35B-A3B Public Release — qwen3-6-35b-a3b is now available to all users

App

Chat

• Video & Audio Attachments in Chat — Attach video and audio files in chat conversations with models that support audio or video input.

• Chat Video Attachments — Chat messages now support multiple video attachments, with the maximum count varying by model.

• Chat Insights & Memoria — Chat insights are now extracted per conversation and combined with Memoria notes

• Image Upload — Redesigned the image upload flow for vision-capable models.

• Automatic Language Detection — Chat now automatically detects and responds in the language of the user's input

• Voice Mode Chat Indicator — Chat surface now visually reflects when voice mode is active.

• Voice Agent Silence Detection — Voice agent now supports silence detection during conversations.

• X Search Toggle — New X Search toggle added to text generation settings, shown when the capability is available

• Audio Reference Clips — Add audio reference clips to prompts using @-mention autocomplete in the composer.

• Drag-and-Drop Reference Videos — Reference videos can now be dragged and dropped directly into chat.

• Image Paste Menu — New paste menu option in chat allows pasting images directly from the clipboard

• Large Context Toggle — New "Large Context" switch in advanced text settings.

• Image Reasoning Toggle — Added a toggle to enable image reasoning for selected models.

• Image Model in Character Chat — Users can now switch the image generation model directly within character chat.

• Resizable Assistant Panel — The assistant panel can now be resized by dragging its edge

• Scroll-to-Bottom Button — Added a scroll-to-bottom button in chat conversations.

• Content Moderation — When a request is blocked by content moderation, a suggested alternative model is now displayed.

• Model Selector Filters — Added sort and filter controls to the all-models view in the model selector

• Model Selector Filtering — The model selector now displays only models the user has access to.

• Recommended Models — Updated the set of Venice-recommended models shown to users.

• Agentic Chat Welcome Message — Agentic chat now displays a welcome message on first open

• Chat Insights Toggle — Chat insights and chat memory can now be enabled or disabled independently

• Agentic Video Acceptance Modal — New confirmation modal to accept or reject output in the agentic video flow.

• Agentic Video Regenerate — Added a regenerate action to agentic video output.

• Auto-Approve Checkbox on Video Card — Added an auto-approve checkbox directly on the agentic video approval card.

• Chat Send Indicator — Unified status indicator showing current chat message state (sending, streaming, complete)

• Agentic Image Moderation — Images generated during agentic workflows now display with a blur and moderation overlay

• Agentic Chat Video Duration — Video generation in agentic chat now defaults to the longest available duration for each model.

• Model Capacity Error — Users now see a "model at capacity" error message when a model is temporarily unavailable instead of a generic failure.

• Content Policy Error Messages — Chat now displays a specific content policy violation error instead of a generic failure message.

• Chat Navigation Button — Replaced the chat preview tab with a chat navigation button

• Video Background Fade — Background video now fades out with an animation when a message is sent

• Delete Chat History Warning — Deleting chat history now displays a warning that Studio and video editor data will also be removed

• Context Usage Meter — The context usage indicator now includes tokens from attached files in its count.

• Context Usage Indicator — Context usage indicator is now hidden by default; a new toggle in Advanced Settings controls its visibility.

• Conditional Sampling Controls — Temperature and Top P sliders are now hidden when the selected model does not support sampling parameters

• Reasoning Mode Tooltip — Added a tooltip explaining reasoning mode when hovering over the toggle

• Agentic Chat Video Descriptions — Rewrote video model descriptions in agentic chat to provide clearer guidance on each model's characteristics and best uses.

• Welcome Copy Update — Removed suggested thought-starters from the chat welcome screen

• Welcome Experience Scope — The onboarding welcome experience now appears only in agent chat

• Public Character Soft Delete — Public characters can now be soft-deleted, removing them from public view while preserving underlying data for potential restoration.

Image, Video and Audio

• Doodle Tool — New freehand drawing tool added to the image editor

• Reference Images — Users can now provide a reference image to guide image generation output.

• Voice Cloning — New voice clone interface added to text-to-speech settings, allowing users to create custom voices.

• Edit with Mask in Enhance — Added mask-based editing (inpainting) to the Enhance workflow

• Style References General Availability — Style References are now enabled for all users, not just Pro

• Image Variant Editing — Edit and combine image variants directly from generated results.

• Seed Audio — Seed Audio now supports multilingual input and word-level timing.

• Reference Images — Added reference image support to image generation

• Style References in Image Generation — Style reference controls added to the Image generation tab

• Voice Picker Redesign — Voice picker reorganized into sections with new audio preview controls for sampling voices.

• Prompt Enhance in Studio — New Prompt Enhance button added to the Studio generation toolbar

• Style Reference Multi-Image — Style references now accept multiple images as input simultaneously

• Image Generation Settings — Added image quality and resolution selectors, plus a toggle to display context token usage

• Audio Timing Templates & Subtitles — Audio Studio now offers timing templates and subtitle file downloads

• Voice Cloning UI — Redesigned voice cloning interface in the Studio with a decoupled layout

• Prompt Optimization Toggle — New toggle in image settings to enable or disable prompt-optimization thinking.

• Image Streaming Toggle — Added a streaming on/off switch to image generation settings

• File Attachments as Media References — File attachments can now be used as media references in generation prompts.

• Image Modal Upscaler — Updated the image modal to support the new upscaler workflow.

• Blocked Image Reason — Blocked images now display the specific block reason instead of downloading a blacked-out image.

• Video Audio Ducking — Playing a video now automatically lowers the volume of other audio

• Composer Media Dropzones — Each slot in the composer now has its own media dropzone and an inline reference shelf for attached references

• Edit Limit Banner — A banner now appears in the editor modal and Studio when the daily edit limit is reached, offering a pay-per-use fallback.

• Image Controls Overlay — Image tab controls now appear in a responsive overlay instead of a fixed panel

• Per-Model Variant Settings — Each model variant in Studio now has its own configurable settings

• Editable Title Colors — Title text colors in the video editor are now editable

• Library Frame Selection — Frame images for video generation can now be picked from the Library

• Auto-Retry on Content Block — Image generation automatically retries when all variants are content-blocked.

• Reference Audio Persistence — Reference audio clips are now saved between sessions and can be downloaded

• Assistant Panel Layout — Assistant panel now docks as a slide-out within the editor layout

• Enhance History Controls — Added visibility controls for prompt enhancement history.

• Assistant Panel Persistence — The assistant panel retains its selected Mind and state when switching between editor tabs

• Assistant Panel Reset — Added a reset button to clear the conversation in the assistant panel

• Variant Model Picker — Model selection for image variants now uses a responsive dialog

• Output Acceptance Thumbnails — Output acceptance modal now shows an input thumbnail and "N of M" counter for multi-variant generations.

• Style Reference Model Selector — Added a model selector within the Style Reference panel

• Custom Voice ID — Added a Custom Voice ID input field in Audio Studio for user-supplied voice identifiers

• Image-to-Video Validation — Image aspect ratio and dimensions are now validated against each video model's requirements before generation

• Reference Video Validation — Reference videos that don't meet format or size requirements are rejected at the time of attachment with an error message.

• Video Upscale Seed — Seed values are now supported for video upscaling, enabling reproducible results.

• Inpaint Credit Tooltip — Added a tooltip showing credit cost for inpaint edits.

• Image Edit Confirmation — Pay-per-use confirmation dialog is now skipped for image edit operations

• Grok Imagine Edit Guidance — Added in-app guidance text for Grok Imagine image editing.

• Video Reference Helper Text — Added instructional text in the video generation UI explaining how to stack multiple video references.

• Model Switcher Video Filters — Improved grouping and filtering of video models in the model switcher.

• Video Studio Retry Button — New retry button to resume polling on failed Video Studio generations

• Auto Aspect Ratio Label — Aspect ratio selector now displays an "Auto" label when automatic aspect ratio is selected

• Video Duration Default — Default duration for video models is now set to the middle option.

• Video Generation Status Animation — Status text during video generation now animates to indicate progress.

• Blurred Source Image During Image-to-Video — A blurred version of the source image is displayed while image-to-video generation is in progress.

• Voice Settings Instructions — Added instructional text to the voice settings screen.

• Likeness Check — Added a "Checking likeness…" placeholder in the likeness verification modal during processing.

• References Tab Empty State — The References tab now displays a placeholder when no references exist

• Image Model Input Limits — Each image model now enforces its actual maximum number of input images

• Style Reference Settings — Style reference controls are now hidden when the selected model does not support them.

• Start Frame Visibility — Start frame option is no longer shown for models that only accept reference images

• Grok Single-Image Edits — Aspect ratio selector is now hidden when performing single-image edits with Grok.

Library and Media

• Save Upscaled Images — Added the ability to save upscaled images directly.

• Reclaim Storage — New user-initiated action to reclaim storage occupied by unused video editor media

• Video Gallery — Improved video gallery layout; inputs now reset automatically after generation completes.

• Header & Media Popovers — Redesigned header and media action popovers

• Copy Last Frame — Added a "Copy Last Frame" button to the fullscreen video viewer.

• Library Banner — Library banner artwork now adapts to the current light or dark theme

• Library Banner Simplification — Simplified the feature banner shown in the Library

• Image Resolution Handling — Updated resolution management in the image viewer

Account and Settings

• Avatar and Cover Image Upload — Users can now upload custom avatar and cover images for their profiles.

• Japanese and Chinese Language Support — Added Japanese and Chinese as available interface languages

• Responsive Header & Autosave — Header adapts to screen size with an overflow menu for actions, and display settings now autosave

• Korean Language — Added Korean to the language switcher.

• Language Switcher — Added new languages to the language switcher.

• Video Background Toggle — Added a setting to disable video backgrounds

• Pro Chat Usage Display — Pro users now see "Unlimited" for remaining chat messages on the dashboard.

• Resend Verification Code — Added a "Resend code" option on the verification screen.

• Pay-Per-Use Image Settings — Added configurable confirmation settings for pay-per-use image generation and a fallback for image-edit attachments

• Usage Limits Display — Added image edit and web scrape actions to the usage limits UI.

• Settings Upgrade Routing — Tapping upgrade prompts in Settings now navigates to the Pro subscription screen.

• User Rate Limits Tab — New tab in settings showing the user's current rate limits

• Inpaint Limits Indicator — Inpaint usage limits are now displayed in settings

• Media Verification Consent — Updated the content and layout of the media verification consent modal.

• API Rate Limits Card — API settings now display a rate limits card showing limits for the user's current tier

• API Key Setup — API settings page now shows base URL and agent prompt configuration more clearly

• Settings Subtitles — Replaced info popovers in Settings with inline subtitle text.

• Language Settings Popovers — Improved language selection popovers in Settings.

• Encrypted Backups — Updated the informational note text on the encrypted backups screen.

Wallet and Payments

• Wallet Payments Without Email — Apple Pay, Google Pay, and Link payments no longer require manually entering an email address

• Custom Credit Purchase — New form allowing users to enter a custom dollar amount when buying credits

• Solana Wallet Top-Up — Solana is now accepted as a payment method for account top-ups and wallet authentication.

• Model Pricing in Selector — Model selector popovers now display contextual per-model pricing.

• Cheapest Model Tag — A "Cheapest" tag now appears in the model switcher to indicate the lowest-cost option.

• Promo Code URLs — Promo codes included in shared URLs are now automatically applied.

• Promo Code Redemption — Promo codes can now be redeemed directly from the Billing settings page

• Paid Image Confirmation — Added a confirmation step before generating images that consume credits.

• Voice Upgrade Dialog — New dialog prompting users to upgrade for voice mode access.

• Resubscribe Button — Added a resubscribe button for Stripe users who previously cancelled their subscription.

• Voice Agent Billing — Voice agent usage is now metered and billed in per-second increments.

• Video Auto-Approve — New auto-approve checkbox and corresponding settings toggle for pay-per-use video generation.

• Cancel Past-Due Subscriptions — Users can now cancel subscriptions that are in a past-due payment state.

• Thinking Surcharge Display — The edit-credit-cost view now shows the per-level surcharge for thinking/reasoning steps

Mobile App

• Agentic Chat on Mobile — Agentic chat call-to-action now appears in the mobile header.

• Image Paste Support — Users can now paste images directly into chat input fields on native mobile.

• Image Quality & Resolution Selectors — Added image quality and resolution selectors to the mobile app, matching the web app.

• Drawer Menu & Folder Creation — Refined drawer menu layout and added a dialog to create new chat folders.

• Multi-Select Delete UX — Improved the flow for selecting and deleting multiple chats at once.

• Long Press Chat Menu — Long-pressing a chat in the sidebar opens a context menu with actions.

• Long Press Image Actions — Long-pressing a generated image now shows a context menu with additional actions.

• Add to Chat — Added an "Add to chat" option when selecting text, with a contextual title label.

• Restore Purchases (iOS) — Added a "Restore Purchases" button to iOS subscription screens for recovering previous in-app purchases

• Billing Settings Screen — Added a new Billing screen to the mobile app's Settings.

• Video Settings — Redesigned video settings with inline descriptions and native-style pickers

• Tablet Chat Layout — Chat screens now enforce a maximum content width on tablet-sized displays

• Usage Settings — Added a Usage section to the mobile app settings

• Drawer Menu Icons — Added icons to navigation drawer menu items.

• Profile Button — Profile button restyled as a floating liquid glass card

• iOS Liquid Glass Headers — Settings headers on iOS now use Apple's native Liquid Glass styling.

• Native iOS Controls in Settings — Settings screens now use native iOS controls instead of custom components.

• Account & Settings — Account and settings screens now use native sheet and stack presentation

• Image Long-Press Menu — Image long-press menu uses liquid glass styling on iOS 26+

• Image Generation Error Messages — Updated content-filter error copy for image generation on mobile.

• Settings Swipe Navigation — Disabled swipe-back gesture in Settings to prevent accidental navigation.

• Tablet Image Sizing — Capped maximum image dimensions on tablet devices.

• iOS Studio Performance — Reduced image upload limits and added frame compression in Studio to lower memory usage on iOS

• Android v1.17.2 — Updated Android APK to version 1.17.2

API

• Voice Cloning API — Added API handlers enabling voice cloning for API clients.

• Reduced API pricing — Reduced API pricing is now available for many models, including DeepSeek V4 Pro, DeepSeek V4 Flash, GLM 4.7 Flash Heretic, MiniMax M2.5, Gemma 4 26B A4B, MiMo V2.5, Kimi K2.7 Code, GLM 5.1, and Qwen3 VL 235B.

• Venice Edit API Resource — Venice Edit is now available as an API-accessible image editing resource.

• Style Reference Endpoint — New API endpoint for applying style references to image generation requests.

• Image API — Added enhance_prompt parameter to the public image generate, edit, and multi-edit API endpoints.

• Reference Audio for Voice Models — Reference audio input is now supported across all voice models.

• Solana Top-Up and Auth — The top-up endpoint and Sign-In with Wallet (SIWX) inference authentication now support Solana wallets.

• Seedance 2.0 Consent API — Added a consent endpoint for Seedance 2.0 video generation via the API.

• Qwen-Edit API Retention — Qwen-Edit remains available in the API; previously planned deprecation has been cancelled.

• Cost Field in LLM Response — New cost property on the LLM response body indicating the request's cost.

• x402 Inference Support — Added x402 protocol support for handling inference payment tasks.

• Agent Tools Endpoint — New GET /api/inference/agent/tools endpoint to list available agent tools.

• Usage History Export — Added an API route to export session-level usage history.

• Usage History API — Added a new billing/usage-history endpoint specification to the API.

• Deprecated Usage Endpoint Rate Limit — The ‘billing/usage’ endpoint is deprecated in favor of /billing/usage-history

• Usage History Endpoint — New usage-history endpoint added for retrieving billing usage data.

• Anthropic Fallbacks Parameter — The API now accepts the Anthropic-style fallbacks parameter on model requests.

• Reasoning Effort in /v1/models — The /v1/models response now advertises reasoning_effort support for compatible models.

• API Credits Order Status — The API credits order status response now includes amount and currency fields.

• Credits Order Status Response — Added amount and currency fields to the API credits order status response.

• Rate Limit Scope — Capability-based rate limits are now applied only to specific endpoints rather than globally.

• Video Content Policy Errors — Improved error responses returned when video generation requests violate content policies.

• Image Edit Parameter — New disable_prompt_optimization_thinking parameter on /v1/image/edit-image to skip prompt optimization.

• Qwen 3.5 9B Rate Limits — Rate limits for qwen3-5-9b doubled from the small-tier defaults.

• Gemma 4 31B Rate Limits — Rate limits for google-gemma-4-31b-it raised to 300 RPM and 6M TPM.

• OpenAPI Spec — /openapi.json now pins the server URL to api.venice.ai.

• Deprecated Usage Endpoint Rate Limit — A 1 RPM rate limit is now enforced on the deprecated billing usage endpoint.

• API Tools Documentation — Added documentation covering API tools usage.

• API Key Generation Guide — Revised the API key generation guide with updated content.

• Rate Limit Docs — Updated the rate limit documentation page.

• API Error Codes Page — Updated the API error codes documentation page.

• Agent CTAs in API Overview — Added agent-related calls to action on the API overview documentation page.

Model Deprecations

• Claude Opus 4.6 Fast — Retired on June 29, 2026.

• Claude Opus 4.7 Fast — Retired on July 24, 2026.

• Qwen Edit 2511 — Retired. API requests using qwen-edit now route to Qwen Edit Uncensored.

• Qwen Edit — Scheduled for retirement.

• LTX Video 2.0 Full Quality — Deprecated. Image-to-video traffic routed to LTX 2 v2.3 Full Image-to-Video.

• LTX Video 2.0 Fast — Deprecated. Image-to-video traffic routed to LTX 2 v2.3 Fast Image-to-Video.

• LTX Video 2.0 19B — Deprecated. Image-to-video traffic routed to LTX 2 v2.3 Full Image-to-Video.

• LTX Video 2.0 19B Distilled — Deprecated. Image-to-video traffic routed to LTX 2 v2.3 Fast Image-to-Video.

• LTX Video 2.0 19B Video-to-Video — Retired on July 22, 2026.

• LTX Video 2.0 19B Distilled Video-to-Video — Retired on July 22, 2026.

• LTX Video 2.0 19B Extend — Retired on July 22, 2026.

• LTX Video 2.0 19B Distilled Extend — Retired on July 22, 2026.

• Runway Gen-4 Aleph — Scheduled for retirement.

Agentic Chat

Agentic Chat is now the default Venice chat experience, with tool
use, media generation, and multi-step workflows available directly
inside a conversation. Users can ask Venice to search, reason, generate
or edit images, create videos, and continue refining the result without
jumping between separate chats.

Open Agentic Chat: https://venice.ai/chat/agent

Here's everything else we shipped.

New Models

The following models have been added to Venice:

Text Models

• Grok Build 0.1 — Text model from xAI. Private. Available to all
  users

• Gemini 3.5 Flash — Google DeepMind's lightweight, low-latency text
  model optimized for speed. Anonymous. Available to all users.

• Qwen 3.7 Max — Large-scale text model from Alibaba Cloud in the
  Qwen 3.7 family. Anonymous. Available to all users.

• Gemma 4 31B Instruct — Google's 31B-parameter open text model with
  instruction tuning. TEE/E2EE. Available to all users.

• Qwen 3.6 35B A3B FP8 — Mixture-of-experts text model from Alibaba
  Cloud with 35B total parameters and 3B active parameters, served in
  FP8 precision for efficient inference. TEE/E2EE. Available to all
  users.

• Gemma 4 26B A4B Uncensored — Uncensored, unfiltered
  mixture-of-experts variant of Google's Gemma 4 with 26B total
  parameters and 4B active parameters. TEE/E2EE. Available to all users.

• Qwen3.6 35B A3B Uncensored — Uncensored, unfiltered variant of
  Alibaba Cloud's Qwen 3.6 mixture-of-experts model with 35B total
  parameters and 3B active parameters. TEE/E2EE. Available to all users.

Image & Video Models

• Grok Imagine High Quality (SOTA) — Image generation model from xAI
  with state-of-the-art quality output. Private. Available to all users.

• Kling V3 Standard Motion Control — Kuaishou's V3 Standard video
  model with motion control support, enabling camera and subject
  movement direction in generated videos. Anonymous. Available to all
  users.

• Kling V3 Pro Motion Control — Kuaishou's V3 Pro video model with
  motion control support, enabling camera and subject movement direction
  in generated videos at higher quality. Anonymous. Available to Pro
  users

Audio Models

• Lyria 3 Pro — Google DeepMind audio and music generation model
  capable of producing high-fidelity instrumental and vocal tracks.
  Anonymous. Available to all users

Model Updates

• Qwen Image Update — Effective June 18, Qwen Image pricing will
  increase from \$0.01 to \$0.03 per generated image. The model will
  also move from height / width parameters to aspect_ratio, with support
  for: 1:1, 3:2, 16:9, 21:9, 9:16, 2:3, 3:4, and 4:5.

• GPT Image 2 Quality Selector — New quality setting added for GPT
  Image 2 image generation.

Web App

New Features & Improvements

• Home Page Redesign — Redesigned home page with updated layout at
  venice.ai/home

• Default Agentic Chat — The default chat route now redirects to
  agentic chat.

• Agentic Chat Message Editing — Messages in agentic chat can now be
  edited in place without resending.

• Auto-Approve Video in Agentic Chat — Video generation requests in
  agentic chat are now auto-approved without requiring manual
  confirmation.

• Drag-to-Folder for Agentic Chat — Agentic chat conversations can
  now be dragged into folders in the sidebar

• Video Studio Enhancements — Added start-from-shared-video,
  advanced model selection, generation queue, save-to-assets, and a
  "taking longer" progress indicator in Video Studio

• Video Download — Videos can now be downloaded directly from the
  video interface

• Download All Videos — New "Download All Videos" button in the
  Studio gallery to batch-download all generated videos.

• In-Browser Camera Capture — Capture photos directly from the
  browser camera in Chat, Video Studio, Image Studio, and multi-edit

• Model Selector Enhancements — Model cards now show an agentic
  filter, code-optimized indicator, capability icons, copy-ID button,
  and context window size.

• Image-to-Video Progress Preview — A blurred version of the source
  image is now displayed while image-to-video generation is in progress.

• Write Code Preset — New "Write Code" preset option added to chat
  for code-focused conversations

• Audio Reference Chips — Audio files can now be attached as
  reference chips via the chat slash menu

• Two-Factor Authentication — Added additional second-factor
  authentication options for account security

• Email Identity Verification — Optional email address can now be
  linked for account identity verification

• Unified Reference Upload — Single upload button that accepts
  image, video, and audio references in one action

• Clickable Creator Name — Creator name in the Social Feed post
  detail header now links to the creator's profile

• Emoji Reactions — Emoji reactions are now available on social
  posts in the Social Feed

• Pinned Featured Models — Featured Image and Video models are now
  pinned and no longer change based on the user's current model
  selection

• Suggested Prompts for Image & Video — Pre-written prompt
  suggestions now appear in the image and video generation interfaces.

• Tool Call UI — New visual indicator in chat showing when the model
  is executing a tool call during a conversation.

• Consumption Limit Reset Period — Consumption limits can now be
  configured with Monthly, Daily, or Total reset periods.

• Image Generation Metadata — Generated images now include
  generation metadata (model, prompt, settings) embedded in the file.

• Image Model Switching — Incompatible settings such as resolution
  or aspect ratio are now automatically reset when switching between
  image models.

• Conversation Mode Minutes Display — Remaining conversation mode
  minutes are now shown in the Storage & Limits settings page.

Wallet and Payments

• Crypto Subscription Credit Display — Credit balance now shown on
  the crypto subscription card.

• Credit Cost Granularity — Credit costs now displayed with
  per-100-character precision.

• Auto Top-Up — New option to automatically replenish credits when
  balance falls below a set threshold.

• Crypto Subscription Management — Upgrade, downgrade, or cancel
  directly from the crypto subscription card.

• Credit Usage History — View a detailed log of past credit usage in
  the wallet section.

• Solana Top-Up Support — Solana is now accepted as a payment method
  for credit top-ups and wallet authentication.

Mobile App

• Agentic Chat — Introduced Agentic Chat (Chat V2), a new multi-step
  chat mode powered by an agent, with access to Venice tools and
  features.

• APK Install Dialog — New dialog prompts Android users to install
  the native APK.

• Referral Screen — New referral screen presented as a modal in the
  mobile app.

• Image Settings — Added pay-per-use configuration for image
  generation in app settings.

• Image Search — Image search results now appear as output items in
  chat.

• Video Carousel — Video carousel now supports Agentic Chat videos.

• Enter Key to Send — Pressing Enter now submits messages in chat.

• Image Modal Aspect Ratio Gating — The image modal now validates
  selected aspect ratios before generation.

• Context Search Rendering — Added inline rendering of context
  search results in chat.

API

• Venice MCP Server — Venice MCP server is now live, available as an
  npm package and on GitHub, with 31 tools covering Chat & Embeddings,
  Image, Video, Audio, and Web capabilities via the Model Context
  Protocol.

• Venice Video MCP — Video generation tools are now available
  through an MCP-compatible workflow for agent builders using Venice
  video capabilities.

• x402 Solana Support — The x402 top-up endpoint and wallet
  inference auth now support Solana.

• Agent Tooling Docs — API docs now include an Agent Tooling section
  covering Venice MCP, Skills, and the Video Harness.

• Private Research Agent Guide — API docs now include a guide for
  building private research agents on Venice.

• Reasoning Effort Options — The API now returns which
  reasoning_effort values each supported model accepts.

• Max Tokens Enforcement — max_tokens is now enforced as the cap for
  the total number of tokens the model generates, including reasoning
  tokens.

• Context-Grep Tool Access — The context-grep tool was enabled for
  public use where supported.

• TTS Response Format Selection — The text-to-speech endpoint now
  accepts a response format parameter to specify the output audio format
  per request.

• Voice Usage Metering — Voice conversation usage tracking switched
  from byte-counting to audio token measurement.

• Voice Conversation Quotas — Voice conversation quotas are now
  enforced during the session instead of only after usage is processed.

• Reference Audio URLs — OpenAPI docs now expose
  referenceaudiourls for supported audio-reference workflows.

• Seedance R2V Audio — Seedance reference-to-video workflows can now
  use reference audio through the public API.

• Video API Docs — API docs now clarify video Base64 behavior and
  remove misleading asset ID references.

• Seedance 2.0 Usage Guide — API docs now include a Seedance 2.0
  guide for API users.

• GPT Image 2 Quality Parameter — GPT Image 2 now supports a quality
  parameter with pricing tied to resolution and quality.

• Venice Edit Multi-Edit Endpoint — Venice Edit now uses a
  multi-edit endpoint, supporting multiple image edits in a single
  request.

• Venice Edit Resource — Venice Edit was added as an API-accessible
  image editing resource.

• Image Edit and Upscale Timeout — Venice image edit and upscale
  requests now support longer processing timeouts.

• Multipart File Uploads — Added a gRPC file chunk streaming client
  for multipart upload workflows.

• Scrape Limits — Added limits for scrape API usage to make web
  extraction behavior more predictable.

• Model Deprecation Fields — New API-only fields on model objects
  indicate deprecation date and replacement model information.

• Deprecation Warning Headers — API responses now include
  deprecation warning headers for affected models.

• Deprecated Model Hiding — Models are automatically removed from
  API model listings once their deprecation date is reached.

• Dynamic Deprecation Table — API docs now generate the model
  deprecation table dynamically.

• Private Models Download Link — API docs now include a private
  models download link.

• Gemma 4 31B API Rate Limits — API rate limits were increased for
  google-gemma-4-31b-it

• Claude API Rate Limits — API rate limits were increased for Claude
  Sonnet 4.6 and Claude Opus 4.6.

• Grok Imagine V2V Pricing — Fixed duration-based pricing for Grok
  Imagine video-to-video usage.

• Music Pricing Granularity — Character-based music pricing now uses
  finer 100-character granularity.

• Burn Read APIs — Added API endpoints for reading token burn data.

• Payments Endpoint — Added a missing payments endpoint used by
  wallet and subscription flows.

• Overload Retry Headers — Overloaded image edit requests now return
  429 with Retry-After where supported.

Model Deprecations

• Qwen 3.5 122B A10B (TEE/E2EE) — Replacement: Qwen 3.6 35B A3B FP8 (TEE/E2EE)

• Grok Imagine Pro — Replacement: Grok Imagine High Quality.

• Grok 4.1 Fast — Replacement: Grok 4.3.

• GLM 5 (TEE/E2EE) — Replacement: GLM 5.1 (TEE/E2EE)

• Qwen 3 Coder 480B — Replacement: Qwen 3 Coder 480B Turbo.

• Kimi K2 Thinking — Replacement: Kimi K2.5.