A4F
v19.4.0•

Gemini 3 Stability Fix, Grok 4 Series & Platform Refinements

Sreejan
SreejanAuthor

Critical stability improvements for Gemini 3.0 and Gemini 2.5 Pro restore full functionality with beta thinking token support. The Grok 4 series debuts on Ultra tier alongside GPT-5.1 expansion across Pro and Ultra tiers. This update also performs strategic catalog optimization with 29 open-source model removals from Provider 2 and resolves image generation timeout issues.

Provider 2

31

Model Additions:

Open Source Image Models Removed (15 models):

Open Source Chat Models Removed (14 models):

Stability Fixes:

  • Gemini 3.0 & Gemini 2.5 Pro: Fixed critical issues causing model failures

    • provider-2/gemini-3-pro-preview - Now fully operational
    • provider-2/gemini-2.5-pro - Stability restored
    • Thinking Tokens (Beta): Added beta support for thinking tokens
      • Tokens may occasionally not display during generation
      • Model responses remain fully functional
      • Complete stability will be addressed in the next update

  • Image Generation Timeout: Increased from 60 seconds to 600 seconds, resolving stuck requests and timeout failures

Provider 3

7

Models Added:

  • Pro Tier:

    • provider-3/gpt-5.1
      • OpenAI's flagship model with 400K context window
      • Advanced reasoning, vision, and function calling support
    • provider-3/gpt-5.1-chat
      • Low-latency instant variant with 272K context window
      • Optimized for rapid conversational responses
      • Full vision and function calling support

  • Ultra Tier Exclusive:

    • provider-3/grok-4-fast
      • xAI's cost-efficient reasoning model with 2M token context
      • Vision and function calling capabilities
    • provider-3/grok-4.1-fast
      • High-throughput model with 2M token context
      • Optimized for parallel processing workloads

Models Removed:

Provider 4

1

Models Added:

Provider 5

2

Models Added:

  • provider-5/gpt-5.1-codex-max - Ultra Tier Exclusive
    • 400K token context window for massive codebases
    • Responses route only (/v1/responses endpoint)
    • Features: verbosity parameter, reasoning, function calling, vision
    • Performance: 3.0-8.0s latency reflecting increased compute
    • Enhanced version of GPT-5.1 Codex with superior code generation capabilities

Feature Enhancement:

  • Verbosity Parameter Expansion: Now available across all GPT 5 and GPT 5.1 series models
    • Supported models: GPT-5, GPT-5 Mini, GPT-5 Nano, GPT-5 Search, GPT-5.1, GPT-5.1 Chat, GPT-5.1 Codex, GPT-5.1 Codex Mini, GPT-5.1 Codex Max
    • Enables fine-grained control over response detail levels
    • Refer to Parameters documentation for usage details

Stability Fixes:

  • Image Generation Timeout: Increased from 60 seconds to 600 seconds, resolving stuck requests and timeout failures

Platform Impact

4
  • Gemini Reliability: Gemini 3.0 and Gemini 2.5 Pro now fully operational with beta thinking token support
  • Ultra Tier Expansion: Grok 4 series and GPT-5.1 Codex Max provide cutting-edge reasoning and coding capabilities
  • Pro Tier Enhancement: GPT-5.1 and GPT-5.1 Chat bring flagship-level capabilities to Pro tier subscribers
  • Catalog Optimization: Removal of 29 open-source models streamlines Provider 2 for production-focused usage

Important Notes

4
  • Thinking Tokens (Beta): Gemini thinking tokens are in beta. Occasional display issues may occur but do not affect model output quality. Full stability coming in the next update.
  • Provider 4 Rate Limits: Gemini 2.5 Flash Image in Provider 4 is limited to 1 image per request across all tiers for stability.
  • Model Alternatives: Removed open-source models may be available through other providers. Check the Models page for alternatives.
  • Verbosity Parameter: Now available on all GPT-5 and GPT-5.1 series models in Provider 5 for enhanced response control.