API Update Announcement

We are pleased to share the latest updates to our API, designed to enhance performance and reliability. Below is a summary of the changes:

• Resolved issue with Google-Embedding-001 model in Provider 3
• Improved functionality of image editing models
• Removed Opus and Opus Thinking features
• Fixed issues with Gemini 2.5 Pro in Provider 1
• Addressed problems with Sonar models in Provider 1
• Corrected token counting issue in Provider 3 during streaming
• Enhanced vision capabilities for select models
• Fixed Mini Max models (40K and 80K)
• Resolved streaming issue in Provider 6

Exciting Provider 6 API Update
We’re thrilled to announce the addition of new models to Provider 6, now available on the free tier for a limited time 👇

• Chat Completion
• Kimi-K2-Instruct
• Llama-4-Maverick
• Llama-4-Scout

• Image Generation
• Flux.1-dev
• Flux.1-pro
• Flux.1.1-pro
• Flux.1-kontext-dev
• Flux.1-kontext-pro
• Flux.1-kontext-max

• Image Editing
• Flux.1-kontext-dev
• Flux.1-kontext-pro
• Flux.1-kontext-max

• Audio Speech
• cartesia/sonic
• cartesia/sonic-2

New Addition:
Added Horizon Beta - OpenAI's upcoming open-source model via OpenRouter, available for free to all users.

Changes to Free Tier:
Removed several models due to high demand affecting paid users' experience. Removed from free tier:

• Text-to-speech models
• Image editing models
• Video generation models

Some models now exclusive to Pro tier only.

Why These Changes:
Current quota insufficient to support free tier demand. Our priority is ensuring paid customers receive uninterrupted service. Cannot provide all premium models for free indefinitely.

Moving Forward:
Will restore free tier models once adequate quota is secured. Thank you for your understanding.

We appreciate your patience as we balance service quality for all users

New Addition:
• Provider 5 now available
- Includes Claude, GPT, and Grok models
- Features Claude 4 and Claude 3.7 with function calling & vision capabilities

Provider Updates:
• Provider 6 modifications:
- Removed some models
- Added new models

• Provider 3 enhancements:
- Updated model lineup

New Model Additions:
• GLM 4.5 & GLM 4.5 Air now available

Bug Fixes:
• Resolved GPT Image 1 issues
• Fixed Qwen Image functionality

Dear Users,

Provider Updates & Current Status:
- Added new providers: Provider 2 and 6 are now active, Provider 5 is working
- Many models removed/updated due to ongoing sustainability issues
- Provider 3 (most stable for Claude/Gemini) remains down until September 7th due to personal reasons
- Open-source models are stable on providers 1, 2, and 6 (direct from official sources)
- OpenAI is stable at Provider 2
- Claude & Gemini are experimental at providers 1 and 5 with limited quota
- Unknown/less renowned providers added for these models - we cannot guarantee stability, model authenticity, or proper routing

Critical Business Challenges:
Our unlimited access model is fundamentally broken. We provide unlimited access but face constant abuse. Claude and Gemini, being closed-source models, are unmaintainable and unstable on our infrastructure.

Everyone on pro/basic plans demands Gemini 2.5 Pro or Claude, with many requesting Opus 4.1 at current pricing - which we cannot sustainably offer. We're barely profitable while handling costly refunds.

Recent issues include 3 DDoS attacks on our servers. We're implementing Cloudflare by September 8th, which may affect speed/latency.

Abuse & Unauthorized Usage:
We've discovered Exoml and other third-party API providers using us as their backend by purchasing multiple pro plans. This widespread abuse is unsustainable.

Immediate Changes:
- Unlimited plans will be removed soon
- New business model launching this month (generous but limited)
- Claude & Gemini guaranteed fixed by September 7th
- No monthly refunds available currently - if unsatisfied by September 8th, refunds will be processed

Personal Note:
I created A4F out of curiosity and hope to help people, but the constant abuse is deeply frustrating. Most users who buy plans abuse the system, forcing these difficult decisions.

We appreciate your patience during this transition to a sustainable model.

Thank you for your understanding.