Major Multi-Provider Update — 54 Models, Provider 5 Rebuild, FLUX 2 Klein, Claude 4.6 & Cloudflare Fix

Critical Fix: Many users were being blocked by Cloudflare when making API requests due to aggressive rate limiting rules. This issue has been fully resolved — all users can now access the API without encountering Cloudflare blocks or captcha challenges.

Major Infrastructure Refresh: Provider 2's entire model catalog has been rebuilt with 54 chat/completion models, updated specifications, new model variants, and refreshed configurations across all families. 28 models available on Free tier, 49 on Basic, and all 54 on Pro/Ultra. Includes 44 models with function calling, 24 with reasoning, and 5 with vision.

Catalog Refresh: 4 new models added and 12 legacy models removed from Provider 3.

4 new image generation models added to Provider 4, including Google's premium Gemini 3 Pro Image and the affordable FLUX 2 Klein family.

The previous Provider 5 (Antigravity) has been decommissioned and a new temporary Provider 5 has been launched in its place.

Provider 7 now fully supports Claude Opus 4.6 with Anthropic's new adaptive thinking mode, replacing the legacy budget-based thinking approach.

• Output Tokens: Increased from 64K to 128K tokens for Claude Opus 4.6
• Adaptive Thinking: Claude 4.6+ models now use adaptive thinking mode, where the model dynamically decides how much reasoning to apply — no manual budget configuration needed
• Legacy Support: Older Claude models continue to use the existing budget-based thinking mode without any changes

• Cloudflare Blocking: Users were being blocked by Cloudflare due to aggressive rate limiting — this has been fully resolved and all users can now access the API without interruption.

• Claude 4.6 Reasoning Not Showing: Reasoning/thinking content from Claude Opus 4.6 was silently dropped in both streaming and non-streaming responses — thinking blocks now correctly flow through to the final response.

• Video Generation Models: Fixed several issues affecting video generation models — more updates and improvements to video generation will be rolling out soon.

• Provider 3 Model ID Correction: The GLM-4.7 free model was registered with an incorrect ID (glm-4.7 instead of glm-4.7-free) — API calls should now use provider-3/glm-4.7-free.

• Provider 2 Refresh: Complete model catalog rebuilt with TEE-secured variants, new model families (Hermes 4.3, Kimi K2.5, Nemotron 3 Nano), and standardized configurations
• Provider 5 Simplification: Rebuilt from a complex 19-model setup to 3 focused, temporary Google Gemini models available on all tiers — permanent Vertex AI integration coming soon
• Image Generation Value: FLUX 2 Klein models offer the most affordable image generation on the platform across all tiers
• Claude 4.6 Reasoning: Full adaptive thinking support ensures users get the most out of Anthropic's latest model capabilities
• GPT-5.2 Access: OpenAI's latest flagship now available on Pro and Ultra tiers through Provider 3
• Cloudflare Resolution: API access restored for all users who were previously experiencing blocks

• Provider 2 Migration: Several models have been renamed with TEE suffixes (e.g., deepseek-r1 → deepseek-r1-tee). Update your integrations to use the new model IDs.
• Provider 5 — Temporary and Unstable: The new Provider 5 is a temporary solution and may be unstable. The previous Provider 5 (Antigravity) was removed because their rate limits were increased beyond what we could sustain. We are actively working on adding official Gemini 3 Pro API support through Google Vertex AI as a permanent, stable replacement.
• Provider 5 Migration: All previous Provider 5 model IDs are no longer valid. Gemini models are now available as provider-5/gemini-3-pro, provider-5/gemini-2.5-flash, and provider-5/gemini-2.5-pro. Claude and OpenAI models should be accessed through Provider 3 or Provider 7.
• FLUX 2 Klein Models: Available across all tiers — the most cost-effective image generation option on the platform.
• Claude Opus 4.6: Adaptive thinking mode is automatic — no configuration changes needed. Output token limit increased to 128K.
• Video Generation: Bug fixes have been applied to video generation models. More improvements and updates to video generation are coming soon.
• Model Availability: Check the Models page for the latest specifications, tier availability, and feature support for all models.