Modèles

Free deterministic test model answered by the gateway itself. Replies with fixed text, or calls the first declared tool, on every protocol, streaming or not. Use it to verify an integration before switching to a real model.

OpenAI's fast, efficient GPT-6.1 model for visual reasoning, precise computer use, and clear, natural writing.

Moonshot AI's flagship 3T-class open-weight model with 1M context, strong coding, 3D creation, and financial research.

Kimi K3.1 Pro served at 2-3x output speed for latency-sensitive workloads.

DeepSeek V4.1 Flash 1017: natively multimodal vision understanding, with stronger coding and agentic performance than the 0910 release, 1M context and 384K max output.

Google's flagship reasoning model for complex software engineering, deep mathematical and scientific problem solving, autonomous multi-step agents, and long decode up to 1M tokens.

Anthropic's fastest and most affordable small model for high-volume, latency-sensitive work such as classification, routing, extraction, and subagents.

Mistral AI's 1.05T-parameter multimodal MoE for reasoning, coding, and agentic work.

OpenAI's upgraded GPT-6 Sol with near-Astra performance for complex coding, computer use, and professional work.

Anthropic's Claude Sonnet 5.5 for everyday coding, agentic tasks, and polished documents, faster than Sonnet 5 at the same price.

MiniMax's frontier multimodal coding model with a 1M context window and tunable thinking depth.

Anthropic's Claude Opus 5.5 for long-running agentic coding and knowledge work, priced below Opus 5 with adaptive thinking always on.

Fireworks' specialized model built on Kimi K3 with shorter reasoning traces for efficient agentic workloads.

OpenAI's cost-efficient GPT-6 model for high-volume workloads such as extraction, summarization, and routing.

OpenAI's frontier GPT-6 model for enterprise agents, coding, writing, and complex knowledge work.

Xiaomi's efficient omni-modal reasoning model (309B total, 15B active) for high-volume coding and agent workflows.

Xiaomi's flagship omni-modal reasoning model (1.02T total, 42B active) for long-horizon agents, coding, and research.

High-speed serving option of GLM-5.3-Flash for coding and agent workflows.

A lower-cost Doubao Seed 2.1 variant for everyday multimodal workloads.

ByteDance's flagship Doubao Seed 2.1 model with 1M context for coding and agent workloads.

Google Gemini 3.8 real-time bidirectional Live API model for low-latency voice agents. Supports text, image, and audio input with text and audio output.

Google Gemini 3.8 Live with background reasoning for real-time voice agents that need deeper thinking. Supports text, image, and audio input with text and audio output.

TypeSafe System One decision model. Evaluates a state against typed questions and returns typed answers (noul probability / choice / score / box) with calibrated confidence; it does not generate text. Image assets (and box questions) require the TypeSafe-Danger-Zone-Images: true header.

DeepSeek V4.1 Flash: natively multimodal vision understanding with 1M context and 384K max output.

OpenAI's fastest image model for high-quality everyday generation and editing, with text and image inputs.

OpenAI's most capable image model for generation and editing where editing precision matters most, with text and image inputs.

OpenAI's most capable model, built for the hardest end-to-end work: complex reasoning, coding, computer use, research, and document creation.

Google's workhorse model for coding and agents, with sharper multi-turn coding, rapid refactoring, code synthesis, and tighter tool-execution logic.

Claude Fable 5.1 extends Claude Fable 5 with stronger long-running agentic coding, knowledge work, and research. Same $10/$50 rates with cache reads at a quarter of the cost.

Tencent Hy4 preview reasoning model served by TokenHub for coding agents, complex tool calls, and multi-step agentic workflows.

Google Gemini 3.5 low-latency speech-to-text over the Live API for bidirectional streaming audio transcription.

Zhipu's first natively multimodal GLM-5 model (320B/18B MoE) with 1M context and visual coding capabilities.

Z.ai's latest flagship reasoning model with a 1M-token context window and controllable thinking length.

Google's most intelligent workhorse model for coding and agentic workflows.

NVIDIA's efficient reasoning model for low-latency agentic workloads.

DeepSeek V4 Flash 0731 weight snapshot with 1M context and 384K max output.

Anthropic's flagship Opus model with enhanced agentic capabilities, priced identically to Opus 4.8.

Claude Opus 5 via third-party reseller channels. Same model family as anthropic/claude-opus-5, served from managed pools with 200K context.

Google's high-throughput Gemini 3.5 model for agentic search, document processing, and translation.

Google's multimodal workhorse model for coding, knowledge work, and agentic workflows.

Moonshot AI's flagship native multimodal model for long-horizon coding, knowledge work, and deep reasoning.

Thinking Machines Lab's general-purpose multimodal model with controllable reasoning effort.

OpenAI's GPT-5.6 alias, routing to GPT-5.6 Sol for frontier reasoning, coding, and production workflows.

A cost-efficient GPT-5.6 model optimized for high-volume reasoning and tool-use workloads.

The frontier GPT-5.6 model for complex reasoning, coding, tool use, and demanding production workflows.

GPT-5.6 Sol via third-party reseller channels (Responses API, 272K tiered billing).

A balanced GPT-5.6 model for strong reasoning and coding performance at a lower price point.

GPT-5.6 Terra via third-party reseller channels (Responses API, 272K tiered billing).

OpenAI's current realtime speech model, responding to audio, image, and text inputs in realtime over WebRTC, WebSocket, or SIP connections.

A cost-efficient version of GPT Realtime 2.1, responding to audio, image, and text inputs in realtime over WebRTC, WebSocket, or SIP connections.

Anthropic's Sonnet model for coding, agents, and complex reasoning with a 1M-token context window.

Claude Sonnet 5 via third-party reseller channels with 200K context.

Z.ai's flagship long-horizon reasoning and coding model with a 1M-token context window.

Anthropic's most capable widely released Mythos-class model for demanding reasoning, long-horizon agentic work, software engineering, knowledge work, vision, and scientific research.

NVIDIA's frontier-scale reasoning model for complex agents, long-context analysis, code, math, and science.