Provider 5 (Nebius) Revocation & New Provider Infrastructure, Provider 2 Image Generation Restored

Platform Discontinuation: Provider 5 powered by Nebius has been completely revoked from the platform. All 41 models have been discontinued effective immediately.

Chat/Completion Models Removed - Free Tier (22 models)

Meta Llama Family:

• provider-5/meta-llama-3.1-8b-instruct-fast - Ultra-fast 8B variant
• provider-5/meta-llama-3.1-8b-instruct - Standard 8B instruction-tuned
• provider-5/llama-guard-3-8b - Safety and content moderation
• provider-5/llama-3.3-70b-instruct - 70B flagship with function calling
• provider-5/llama-3.3-70b-instruct-fast - Fast variant of 70B flagship

NVIDIA Models:

• provider-5/nemotron-nano-v2-12b - NVIDIA's efficient 12B Nemotron Nano V2

Google Gemma:

• provider-5/gemma-2-2b-it - Compact 2B instruction-tuned
• provider-5/gemma-2-9b-it-fast - Fast 9B Gemma 2 variant
• provider-5/gemma-3-27b-it - 27B instruction model with function calling
• provider-5/gemma-3-27b-it-fast - Fast variant with function calling

Qwen Series:

• provider-5/qwen2.5-coder-7b-fast - Fast coding specialist
• provider-5/qwen3-32b - 32B flagship with function calling
• provider-5/qwen3-32b-fast - Fast 32B variant
• provider-5/qwen3-30b-a3b-instruct-2507 - 30B MoE with function calling

DeepSeek Reasoning:

• provider-5/deepseek-r1-0528 - R1 reasoning model
• provider-5/deepseek-r1-0528-fast - Fast R1 variant

Zhipu AI:

• provider-5/glm-4.5-air - Efficient GLM 4.5 with reasoning

OpenAI Open-Source:

• provider-5/gpt-oss-120b - 120B open-source GPT
• provider-5/gpt-oss-20b - 20B efficient open-source GPT

Moonshot AI:

• provider-5/kimi-k2-instruct - K2 instruction-tuned model

NousResearch:

• provider-5/hermes-4-70b - 70B Hermes 4 with function calling

PrimeIntellect:

• provider-5/intellect-3 - Intellect 3 reasoning model

Chat/Completion Models Removed - Basic Tier (11 models)

• provider-5/llama-3.1-nemotron-ultra-253b - NVIDIA's 253B ultra-scale model
• provider-5/qwen3-235b-a22b-instruct-2507 - 235B flagship
• provider-5/qwen3-235b-a22b-thinking-2507 - 235B thinking variant
• provider-5/qwen2.5-vl-72b-instruct - 72B vision-language model
• provider-5/deepseek-v3-0324 - DeepSeek V3 with function calling
• provider-5/deepseek-v3-0324-fast - Fast DeepSeek V3 variant
• provider-5/glm-4.5 - Full GLM 4.5
• provider-5/kimi-k2-thinking - K2 with enhanced reasoning
• provider-5/qwen3-30b-a3b-thinking-2507 - 30B thinking variant
• provider-5/qwen3-coder-30b-a3b-instruct - 30B coding specialist
• provider-5/qwen3-next-80b-a3b-thinking - 80B next-gen thinking model

Chat/Completion Models Removed - Pro Tier (2 models)

• provider-5/qwen3-coder-480b-a35b-instruct - 480B flagship coding model
• provider-5/hermes-4-405b - 405B NousResearch flagship

Embedding Models Removed (4 models)

• provider-5/bge-en-icl - BAAI 4096-dimension embeddings
• provider-5/bge-multilingual-gemma2 - BAAI multilingual embeddings
• provider-5/e5-mistral-7b-instruct - Intfloat embeddings
• provider-5/qwen3-embedding-8b - Qwen embeddings

Image Generation Models Removed (2 models)

• provider-5/flux-dev - Black Forest Labs development model
• provider-5/flux-schnell - Black Forest Labs fast generation

New High-Performance Provider: A new Provider 5 infrastructure has been integrated with 13 premium models spanning Google Gemini, Anthropic Claude, and OpenAI families.

Google Gemini Models (9 models)

Free Tier:

• provider-5/gemini-2.5-flash-lite - Free, Basic, Pro, Ultra

  • Lightweight and cost-effective variant optimized for high-volume tasks
  • Context Window: 524K (Free) / 786K (Basic) / 1M (Pro/Ultra)
  • Output Tokens: 65,536
  • Features: vision, function_calling

• provider-5/gpt-oss-120b-medium - Free, Basic, Pro, Ultra

  • OpenAI's balanced open-weight reasoning model optimized for medium-complexity tasks
  • Context Window: 64K (Free) / 90K (Basic) / 128K (Pro/Ultra)
  • Output Tokens: 32,768
  • Features: reasoning, vision, function_calling

Basic Tier:

• provider-5/gemini-2.5-flash-thinking - Basic, Pro, Ultra

  • Enhanced reasoning variant of Gemini 2.5 Flash with chain-of-thought capabilities
  • Context Window: 256K (Basic) / 512K (Pro) / 1M (Ultra)
  • Output Tokens: 65,536
  • Features: reasoning, vision, function_calling

• provider-5/gemini-2.5-flash - Basic, Pro, Ultra

  • Google's best model for balancing reasoning and speed with 1M token context
  • Context Window: 256K (Basic) / 512K (Pro) / 1M (Ultra)
  • Output Tokens: 65,536
  • Features: reasoning, vision, function_calling, audio

Pro Tier:

• provider-5/gemini-2.5-pro - Pro, Ultra
  • Google's most capable model with advanced reasoning and 1M token context
  • Context Window: 1M (Pro/Ultra)
  • Output Tokens: 65,536
  • Features: reasoning, vision, function_calling, audio

Ultra Tier Exclusive:

• provider-5/gemini-3-pro-low - Ultra Only

  • Gemini 3 Pro Low - Google's efficient variant for faster inference
  • Context Window: 1M (Ultra)
  • Output Tokens: 65,536
  • Features: reasoning, vision, function_calling

• provider-5/gemini-3-pro-high - Ultra Only

  • Gemini 3 Pro High - Google's premium variant with maximum quality
  • Context Window: 1M (Ultra)
  • Output Tokens: 65,536
  • Features: reasoning, vision, function_calling

• provider-5/gemini-3-pro-image - Ultra Only

  • Gemini 3 Pro Image - Vision-enabled model for image understanding and analysis
  • Context Window: 1M (Ultra)
  • Output Tokens: 65,536
  • Features: vision

• provider-5/gemini-3-flash - Ultra Only

  • Google's fast and efficient multimodal model
  • Context Window: 1M (Ultra)
  • Output Tokens: 65,536
  • Features: reasoning, vision, function_calling, audio

Anthropic Claude Models (4 models)

Ultra Tier Exclusive:

• provider-5/claude-sonnet-4-5 - Ultra Only

  • Claude Sonnet 4.5 - Advanced reasoning model with excellent performance across diverse tasks
  • Context Window: 200K (Ultra)
  • Output Tokens: 64,000
  • Features: vision, function_calling

• provider-5/claude-sonnet-4-5-thinking - Ultra Only

  • Claude Sonnet 4.5 Thinking - Extended reasoning variant with enhanced chain-of-thought
  • Context Window: 200K (Ultra)
  • Output Tokens: 64,000
  • Features: reasoning, vision, function_calling

• provider-5/claude-opus-4-5 - Ultra Only

  • Claude Opus 4.5 - Anthropic's most powerful model with exceptional reasoning and creative capabilities
  • Context Window: 200K (Ultra)
  • Output Tokens: 64,000
  • Features: reasoning, vision, function_calling

• provider-5/claude-opus-4-5-thinking - Ultra Only

  • Claude Opus 4.5 Thinking - Anthropic's most powerful reasoning model with extended chain-of-thought
  • Context Window: 200K (Ultra)
  • Output Tokens: 64,000
  • Features: reasoning, vision, function_calling

Service Restoration: All Provider 2 image generation models have been fully restored and are now operational after recent infrastructure issues.

• Full Functionality: All image generation endpoints are now working correctly
• Model Availability: All previously available image generation models are accessible
• Stability: Infrastructure issues have been resolved ensuring consistent performance
• No Action Required: Users can resume image generation workflows immediately

This fix addresses the issues that were affecting Provider 2 image generation capabilities, ensuring reliable access for all users.

Upcoming Feature: New video generation models are expected to be added to Provider 2 very soon.

• Timeline: Expected deployment within the next day
• Provider: Video generation will be available through Provider 2
• Details: Full model specifications will be announced upon release

Stay tuned for the official announcement with complete details on the new video generation capabilities.

Provider 5 New Infrastructure Breakdown:

• Google Gemini Models: 9 models with vision, function calling, reasoning, and audio capabilities
• Anthropic Claude Models: 4 models with advanced reasoning and vision support
• Free Tier Access: 2 models available on Free tier (Gemini 2.5 Flash Lite, GPT OSS 120B Medium)
• Ultra Tier Premium: 8 exclusive models including Gemini 3 series and Claude Opus 4.5

Feature Distribution:

• Vision: 13 models with image analysis support
• Function Calling: 12 models with tool/function calling support
• Reasoning: 10 models with enhanced reasoning capabilities
• Audio: 3 models with audio processing capabilities

• Provider Transition: Nebius infrastructure fully replaced with new high-performance provider
• Premium Model Access: Ultra tier users gain access to cutting-edge Gemini 3 and Claude Opus 4.5 models
• Image Generation Reliability: Provider 2 image generation fully restored for uninterrupted creative workflows
• Free Tier Value: 2 high-quality models remain accessible on Free tier despite provider transition
• Video Generation Preview: New video generation capabilities coming soon to expand platform offerings

• Provider 5 Migration: All users of the previous Provider 5 (Nebius) models should migrate to alternative providers. Model IDs have changed completely.
• Embedding Models: No embedding models are available in the new Provider 5. Use Provider 2 or other providers for embedding needs.
• Image Generation: No image generation models in new Provider 5. Use Provider 2 for image generation (now fully restored).
• Ultra Tier Value: 8 of the 13 new models are Ultra-exclusive, providing significant value for Ultra subscribers.
• Video Generation: Watch for the upcoming announcement on new video generation capabilities expected tomorrow.