Understanding AI Models and Providers
Nuveki routes to models from several providers, each with different strengths. Picking the right model for a task can noticeably change the quality, speed, and cost of a response. You set a model per agent, and you can switch at any time.
Claude (Anthropic, via AWS Bedrock)
- Fable 5 — the most capable model on the platform. Reach for it on the hardest reasoning, analysis, and long-form writing. Premium tier.
- Opus 4.8 — top-tier reasoning for complex analysis, nuanced writing, and multi-step problems. Premium tier.
- Sonnet 5 — the balanced default: strong at coding, analysis, and conversation, fast enough for everyday use.
- Haiku 4.5 — the fastest Claude. Ideal for quick answers, simple tasks, and high-volume use.
OpenAI
- GPT-5.6 (Sol, Terra, Luna) — OpenAI's latest general-purpose family: Sol is the flagship, Terra a balanced everyday model, and Luna a fast, affordable tier.
- GPT-5.5 Pro — OpenAI's premium reasoning flagship; premium tier.
- GPT-5.4 family (GPT-5.4, Pro, Mini, Nano) — a range from capable down to ultra-fast and inexpensive.
- O3 / O3 Pro — reasoning-focused models tuned for math, logic, and structured problems. Premium tier.
Google (Gemini)
- Gemini 3.1 Pro — high-capability model with strong multimodal understanding.
- Gemini 3.5 Flash / 3 Flash / Flash-Lite — fast, efficient models for quick interactions.
xAI (Grok)
- Grok 4.3 — fast, general-purpose conversational model.
- Grok 4.5 — Grok's flagship reasoning model for harder problems.
- Grok 4.20 Multi-Agent — runs several agents in parallel for deep research.
Open-weight models (via Bedrock)
For cost-sensitive or specialized workloads there are open-weight options too: DeepSeek V3.2, Qwen3 (several sizes, including a coding-focused variant), MiniMax M2.5, GLM 5.2, and Kimi K2.5.
Choosing a model
- Deep reasoning, nuanced writing → Fable 5 or Opus 4.8
- Everyday work, coding, analysis → Sonnet 5 or GPT-5.6 Sol
- Speed and high volume → Haiku 4.5, a Gemini Flash model, or a Mini/Nano tier
- Math and structured logic → O3 or Grok Reasoning
- Coding with lots of context → Grok 4.5 or a Qwen3 coder model
Model access by tier
- Free — fast, efficient models: Haiku, Sonnet, the Gemini Flash family, Grok 4.3, a GPT Mini, plus open-weight options (DeepSeek, Qwen3, MiniMax).
- Starter — adds higher-capability standard models: Gemini 3.1 Pro, Grok Reasoning, GPT-5.4 and the GPT-5.6 family, and more open-weight models.
- Pro & Team — every model, including the premium flagships: Fable 5, Opus 4.8, O3 and O3 Pro, and the GPT Pro tiers.