A4F
v19.6.0•

New Provider 5 Launch & Provider 8 Tier Optimization

Sreejan
SreejanAuthor

This major infrastructure update introduces a new high-performance infrastructure as Provider 5 with 41 cutting-edge models while retiring the previous legacy provider. Provider 8 receives comprehensive tier optimization with standardized context window allocation following the 50%/70%/100%/100% distribution rule across free/basic/pro/ultra tiers. Strategic tier restrictions enhance platform sustainability while maintaining broad model access.

⚠️ Provider 5 (Legacy) - Complete Retirement

82

Platform Discontinuation: Legacy Provider 5 has been completely removed from the platform. All 82 models have been discontinued.

Image Generation Models Removed (14 models)

Chat/Completion Models Removed (45 models)

OpenAI GPT Series:

OpenAI GPT-3.5 Series:

OpenAI GPT-4.1 Series:

OpenAI GPT-5 Series:

OpenAI GPT-5.1 Series:

OpenAI o-Series:

Audio Models Removed (18 models)

Embedding & Moderation Models Removed (5 models)

Provider 5 - New Infrastructure Launch

41

New Infrastructure Integration: A new high-performance infrastructure joins as Provider 5 with 41 high-performance models spanning chat, reasoning, embedding, and image generation capabilities.

Chat/Completion Models - Free Tier (22 models)

Meta Llama Family:

NVIDIA Models:

Google Gemma:

Qwen Series:

DeepSeek Reasoning:

Zhipu AI:

  • provider-5/glm-4.5-air - Free Tier
    • Efficient GLM 4.5 with reasoning and function calling
    • 128K context window

OpenAI Open-Source:

Moonshot AI:

NousResearch:

PrimeIntellect:

Chat/Completion Models - Basic Tier (11 models)

Chat/Completion Models - Pro Tier (2 models)

Embedding Models (4 models) - All Tiers

Image Generation Models (2 models) - All Tiers

  • provider-5/flux-dev - Free Tier

    • Black Forest Labs development model
    • Max Images: 1/2/4/4 per request (Free/Basic/Pro/Ultra)

  • provider-5/flux-schnell - Free Tier

    • Black Forest Labs fast generation
    • Max Images: 1/2/4/4 per request (Free/Basic/Pro/Ultra)

Provider 8 - Tier Optimization & Context Window Standardization

31

Infrastructure Update: Provider 8 models now follow standardized context window allocation with the 50%/70%/100%/100% distribution rule for free/basic/pro/ultra tiers. Strategic tier restrictions applied for resource optimization.

Free Tier Expansions (2 models)

Ultra Tier Restrictions (7 models)

Chat/Completion Models:

Image Generation Models:

Pro Tier Restrictions (3 models)

Basic Tier Restrictions (5 models)

Context Window Adjustments (14 models)

The following models received context window standardization without tier changes:

Context Window Distribution Rule

1

Standardized Allocation: All Provider 5 and Provider 8 models now follow consistent context window distribution across tiers:

  • Free Tier: 50% of maximum context window
  • Basic Tier: 70% of maximum context window
  • Pro Tier: 100% of maximum context window
  • Ultra Tier: 100% of maximum context window

This standardization ensures predictable behavior across all models while optimizing resource allocation for platform sustainability.

Platform Impact

5
  • New Provider Infrastructure: New Provider 5 integration brings 41 high-quality models with excellent uptime and performance characteristics
  • Free Tier Enhancement: 22 chat models and 2 image generation models now available on Free tier through Provider 5
  • Embedding Capabilities: 4 new embedding models provide comprehensive text embedding options for RAG applications
  • Premium Model Curation: Strategic tier restrictions for Gemini 3, Claude 4.5, and Grok 4.1 ensure sustainable access for premium subscribers
  • Multi-Provider Redundancy: Many removed models now available through the new Provider 5 infrastructure with enhanced stability

Important Notes

6
  • Provider 5 Migration: Users of previous Provider 5 models should migrate to the new Provider 5 or alternative providers. Model IDs have changed.
  • Context Window Changes: Free and Basic tier users will experience reduced context windows following the 50%/70% allocation rule. Plan usage accordingly.
  • Premium Model Access: Claude Sonnet 4.5, Gemini 3 Pro, and Grok 4.1 are now Ultra-exclusive. Use Provider 7 for Claude Sonnet 4.5 with broader tier access.
  • Image Generation Limits: Premium image models (Nano Banana Pro, Seedream 4.5, GPT Image 1.5) restricted to Ultra tier. Flux models in Provider 5 offer free alternatives.
  • Embedding Models: All 4 new embedding models are available across all tiers with competitive pricing.
  • GLM Series: GLM 4.7 and thinking variants now require Pro tier. GLM 4.6 series requires Basic tier minimum.