Back to blog

Q1 2026: Video Gen, Image Studio & Enterprise Features

Three months of updates: video generation, Image Studio, sessions, GPT-5.4 family, enterprise guardrails, 5+ new providers, and much more.

Q1 2026 Feature Roundup

Here's everything new in LLM Gateway over the first quarter of 2026 — from video generation and image editing to enterprise security features and a wave of new models.

Video Generation

LLM Gateway now supports video generation — submit text prompts and receive generated videos via an async job-based API. Jobs are tracked with status polling and proper error handling for failed generations.

Image Studio & Image APIs

Image Studio

The Playground gained a dedicated Image Studio at /image — a gallery-based UI purpose-built for image generation:

  • Compare mode — Select up to 3 image models and generate in parallel, with side-by-side results
  • Full prompt controls: aspect ratio, resolution, pixel dimensions, image count (1–4)
  • Scrollable gallery, click-to-zoom, hover-to-download
  • Recent prompts saved in sidebar for quick re-use

OpenAI-Compatible Image APIs

Two new endpoints for programmatic image work:

  • /v1/images/generations — Generate images with an OpenAI-compatible request format
  • /v1/images/edits — Edit existing images by sending an image + prompt describing changes
  • Both support aspect_ratio parameter and resolution-based pricing

AI SDK Provider: generateImage()

The @llmgateway/ai-sdk-provider now supports Vercel AI SDK's generateImage() via llmgateway.image().

Sessions & Activity

Sessions let you group related requests into trackable conversations:

  • Filter activity logs by session
  • View streaming cost breakdowns per session
  • Improved activity log filters across the board

Enterprise Features

Guardrails

Enterprise-grade guardrails with configurable rules and actions for content moderation, safety policies, and compliance requirements.

Audit Logs

Full audit logging across all organizations — every significant action is tracked with timestamps, actors, and details. Reads are gated on enterprise plans.

Enterprise Contact API

Enterprise contact form submissions now route through the API for better tracking and follow-up.

Automatic Retry & Fallback

When a provider fails (5xx, timeout, network error), the gateway now automatically retries on the next best provider — transparently, within the same API call:

  • Up to 2 retries before returning an error
  • Full routing transparency with routing array in metadata
  • Retried logs linked to their successful replacements
  • X-No-Fallback: true header to opt out

New Models

GPT-5.4 Family

  • GPT-5.4 — 1.05M context, $2.50/$15.00 per 1M tokens, reasoning support
  • GPT-5.4 Pro — 1.05M context, $30.00/$180.00 per 1M tokens, more compute for harder problems
  • GPT-5.4 Mini — Smaller, cheaper variant
  • GPT-5.4 Nano — Lightest variant for high-volume workloads

GPT-5.2/5.3 Codex & Chat

New Codex and Chat variants for the GPT-5 family added with Bedrock and direct mappings.

Gemini Updates

  • Gemini 3.1 Flash Image Preview — Fast image generation with 0.5K–4K resolutions
  • Gemini 3 Pro Preview deprecated in favor of newer models
  • Unified Grok-4-fast and Grok-4-1-fast models with reasoning-based routing

Other Models

  • MiniMax M2.7 mappings
  • Moonshot reasoning option enabled
  • XAI reasoning output configuration
  • o4-mini added to OpenAI models

New Providers

  • EmberCloud — New provider expanding routing options
  • Together AI — Access to Together's model catalog
  • AWS Bedrock — Extended mappings and cache pricing fixes
  • CanopyWave — Cache pricing added (marked unstable due to availability)

Billing & Payments

  • Stripe Checkout for credits top-up — streamlined payment flow
  • Gift credits functionality added to dashboard
  • Payment method validations for better error handling
  • User-friendly error messages on card decline (no more 500s)
  • Resolution-based image pricing for accurate cost tracking

UI & Design

  • Redesigned landing page sections
  • Redesigned Code app with fresh UI
  • Revamped cost simulator and navigation
  • Model category pages — Browse models by category
  • Status page indicator — Quick link from dashboard
  • SEO audit improvements and canonical URLs

Platform & Reliability

  • Reasoning max tokens — Control reasoning depth with configurable limits
  • Streaming cost tracking — Real-time cost calculation during streaming responses
  • Cached token accounting — Fixed double-counting in storage costs and Bedrock streaming
  • Content filter classification — ByteDance and xAI content safety blocks properly handled
  • Bedrock tool schema sanitization — Cleaner tool calling with AWS models
  • IAM enforcement — deny_providers rules correctly enforced during routing
  • Follow-up lifecycle emails — Automated email sequences for user onboarding

Documentation & Guides

  • Codex CLI guide — Using LLM Gateway with OpenAI Codex CLI
  • Autohand guide — Autohand integration walkthrough
  • Clawdbot guide — Clawdbot integration guide
  • Migration guides — New migrations section covering LiteLLM and others
  • Themed screenshots — Light/dark mode screenshots across learn section
  • Introduction page — New docs landing page

Admin Dashboard

  • History graphs and cost charts with date pickers
  • Model detail pages with performance data
  • Members tab on organization detail
  • Revenue tracking with proper transaction filtering
  • Log cards matching UI activity format

Explore all models | Try the Playground | Get started now