33 active models · updated July 11, 2026
Zinley is one agent that runs on a team of frontier models. For every task, Auto picks the model that fits — deep models for hard reasoning, fast models for quick moves. You can also pin any specific model whenever you want. This page is the full active lineup.
The lineup evolves: when a better model ships, we add it, route new work to it, and retire what it replaced — you never have to migrate anything yourself.
The backbone of Zinley — deep reasoning, careful tool use, and long autonomous runs.
Claude Fable 5
Anthropic
Deepest reasoning for the hardest, longest tasks
Claude Opus 4.8
Anthropic
Heavyweight reasoning with fast output
Claude Opus 4.7
Anthropic
Previous Opus generation, still available to pin
Claude Opus 4.6
Anthropic
Previous Opus generation, still available to pin
Claude Opus 4.5
Anthropic
Previous Opus generation, still available to pin
Claude Sonnet 5
Anthropic
Sharp and quick, excellent with tools
Claude Sonnet 4.5
Anthropic
Balanced speed and depth for daily tasks
Claude Sonnet 4
Anthropic
Earlier Sonnet generation, still available to pin
Claude Sonnet 3.7
Anthropic
Earlier Sonnet generation, still available to pin
Claude Haiku 4.5
Anthropic
Instant answers and light tasks — free for everyone
OpenAI’s frontier line, served from Azure AI Foundry — the GPT-5.6 family runs a million-token context.
GPT-5.6 Luna
OpenAI
The everyday frontier default — fast, sharp, on every plan
GPT-5.6 Terra
OpenAI
Deep reasoning at full context
GPT-5.5
OpenAI
Heavy reasoning and orchestration work
GPT-5.6 Sol
OpenAI
The flagship of the 5.6 family
Meta’s frontier reasoning line, fresh from the Meta Model API — a deep-work specialist you can switch on for heavy tasks.
Muse 4.1
Meta
Deep reasoning and heavy agent work — enable it in Settings → Speed & Intelligence
Google’s multimodal line — from long-context analysis down to near-instant replies.
Gemini 3.1 Pro Preview
Long-context multimodal analysis
Gemini 3.5 Flash
Fast multimodal responses
Gemini 3.1 Flash Lite
The lightest, quickest jobs
The best open models, served exclusively from US infrastructure — the original developers never receive your data.
Kimi K2.7 Code
Moonshot AI
Elite coding and agentic tool use
Qwen 3.7 Plus
Qwen (Alibaba)
Strong all-rounder with a big context window
GLM-5.2
Zhipu AI
Reasoning-heavy tool work
MiniMax M3
MiniMax
Fast, economical agentic runs
DeepSeek V4 Pro
DeepSeek
Million-token context for huge codebases and documents
Gemma 4 31B
Google DeepMind
Compact open model for quick jobs
The creative and perception side of the lineup.
Nano Banana Pro
Highest-quality image generation and editing
Nano Banana 2
The default image model — fast and reliable
Nano Banana Lite
Quick, economical image work
GPT Image 2
OpenAI
Alternate image style; automatic fallback
GPT Image 1.5
OpenAI
Alternate image style; automatic fallback
Veo 3.1
Video generation with audio
Veo 3
Video generation, standard quality
Veo 3 Fast
Quick video drafts
GPT-4o Transcribe
OpenAI
Speech-to-text for calls and voice notes
Every model in this lineup runs under the same rules: your data is never used to train any model, and frontier providers (Anthropic, OpenAI, Google) operate under zero data retention agreements. Open-weights models — Kimi, Qwen, GLM, MiniMax, DeepSeek, Gemma — are served exclusively from US-based infrastructure; the labs that originally built them never receive your data. Full details on the Security page.
Every plan runs on this lineup — higher plans unlock more models to hand-pick and higher limits, and on every paid plan Auto runs your tasks on frontier Claude. Whatever the model, everything is billed the same way: credits, explained on Usage & credits.