AIHubMix has released new agent-readable access surfaces so coding agents can autonomously discover and invoke models from its gateway. The package includes an agents.md onboarding guide that agents execute directly, per-model llms.txt files at fixed URLs (e.g., https://aihubmix.com/model/{model_id}/llms.txt) covering The release addresses the failure modes agents hit when guessing model IDs, parameters, and pricing from training memory: stale IDs, unsupported parameters, and outdated costs. Agents are instructed to fetch live data and verify by inspecting the response body and echoed model, not just HTTP 200. The Gateway exposes 50

AIHubMix is offering OpenAI's GPT-5.6 Sol at 50% off via a promotional model ID (gpt-5.6-sol-disc), matching OpenRouter's discount at $2.50 per million input tokens and $15 per million output tokens for requests up to 272K tokens. While the headline rates are identical, AIHubMix does not add a separate platform fee, wh A worked example of 1M input and 1M output tokens shows $17.50 on AIHubMix versus approximately $18.46 on OpenRouter after the platform fee. The discounted route is exposed through OpenAI-compatible Chat Completions and Responses APIs with streaming support, unlike OpenAI's Batch and Flex options that trade latency for

AIHubMix announced that free tiers of OpenAI's GPT-5.5 and GPT-Image-2 are now available on its OpenAI-compatible gateway, with no credit card or trial expiry, joining 27+ other free models including Gemini 3, GLM-5.1, Kimi, and Xiaomi MiMo. AIHubMix subsidizes the inference cost because OpenAI's official API does not GPT-5.5-free positions itself as OpenAI's most capable available model, with reasoning, agent orchestration, tool use, code generation, and data analysis upgrades, while GPT-Image-2-free is the first OpenAI image model with built-in reasoning and 99% character-level text rendering accuracy across Latin, CJK, and Hindi

AIHubMix's changelog shows a rapid-fire model and feature rollout through July and early August 2026. July highlights include 30 new models added (claude-opus-5, GPT-5.6, kimi-k3, qwen3.8-max-preview), media and 3D generation endpoints, an MCP Server, and an 80% official price cut on gpt-5.6-luna (now $0.20 → $1.20) al August 2026 additions tracked in the changelog include deepseek-v4-pro-0813, grok-4.6, qwen3.8-2.4t-a95b, nemotron-lightning-3.5-30b-a3b, muse-glimmer-30b, the agnes-2.5 image and text family, wan3.0-video, doubao-seedance-2-5-260628, qwen-image-3.0 / 3.0-pro, qwen3.8-max, and minimax-h3 video. A new Structured Output

AIHubMix's changelog dated August 6, 2026 documents the addition of two Alibaba Qwen image models to its unified API. qwen-image-3.0 supports text-to-image generation, reference-image creation, and image editing, positioned for social media, e-commerce, creative design, and high-frequency content production with a bala The same changelog entry covers a Gemini embedding upgrade: gemini-embedding-2-preview now accepts text, image, audio, video, and document inputs through both the OpenAI-compatible embeddings endpoint and the native Gemini path, enabling multimodal search, content understanding, and knowledge-base applications within a