External APIs
The external APIs used as overflow to the local core, alongside sovereign models, to absorb spikes or access on-demand capacity.
OpenAI (ChatGPT)
OpenAI offers the ChatGPT family and GPT models as a SaaS, leading the field in reasoning, code and multimodal capability.
Anthropic (Claude)
Anthropic offers the Claude family as a SaaS, recognised for long-form reasoning, reliable code and safety.
Amazon Bedrock
Amazon Bedrock aggregates models from several providers as a SaaS, including Anthropic, Meta, Mistral and Amazon, behind a single managed API.
OpenRouter
OpenRouter aggregates hundreds of models as a SaaS behind a single interface, including the leading open Chinese models (DeepSeek, GLM, Qwen, Kimi),…
Azure AI Foundry
Azure AI Foundry, Microsoft's unified platform, provides SaaS access to a broad model catalogue, including GPT from OpenAI, Llama, Mistral and Phi,…
DeepSeek (API)
DeepSeek serves V4.1 Flash 1017 under the deepseek-flash name, at $0.15 and $0.60 per million off-peak, with a 1M context and open weights.
Mistral AI (API)
Mistral hosts in the European Union by default, the US endpoint being an explicit choice. Mistral Large 4.0 costs $0.68 and $2.09 per million.
Alibaba (Qwen)
Alibaba serves the Qwen range through Model Studio. Qwen3.8 Max Prime costs $4 and $12 per million, and the weights of several Qwen models are published.
Z.ai (GLM)
Z.ai serves the GLM range. GLM 5.3 Prime costs $2.80 and $8.80 per million, and the editor publishes no retention period for API inputs.
Google (Gemini)
Google serves Gemini 4 Argon at $2 and $10 per million until 30 November 2026, then double. The free tier trains the models.
Meta (Muse)
Meta serves Muse Wonder at $4 and $20 per million. Muse Wonder Full has no per-token price and is obtained by subscription.
xAI (Grok)
xAI serves Grok 4.7 at $2 and $6 per million. The editor keeps thirty days for abuse auditing, without training on inputs.