AI Models
Guide

The versatile all-rounder with dynamic internal routing.

• GPT-5.4 is the current flagship with a 1M token context window
• Uses an internal router to select the right sub-model per request in real time
• Native computer use and tool calling for agentic automation
• Strong at documentation, unit tests, and complex SQL queries

The developer favorite for coding, reasoning, and safety.

• Opus 4.6 (flagship), Sonnet 4.6 (best value), Haiku 4.5 (fast/cheap)
• Leads human-preference leaderboards; strong ARC-AGI-2 scores
• Agentic capabilities: autonomous multi-step coding tasks
• Known for safety, steerability, and high-quality long-form writing

Multimodal powerhouse with top benchmark breadth.

• Gemini 3.1 Pro leads the Artificial Analysis Intelligence Index
• Native multimodal: processes text, images, audio, and video natively
• 1M token context window; deep Google Workspace integration
• Strong value at $2/$12 per million input/output tokens

Real-time data meets raw reasoning power.

• Grok 4.20 uses native multi-agent architecture with four specialist agents that debate before responding
• Real-time integration with X (Twitter) for current events
• Scored 100% on AIME 2025 math competition (Heavy variant)
• Less filtered personality; positioned as an alternative to corporate AI

Search-native AI built for grounded, cited answers.

• Built on Llama 3.3 70B, further trained for search-grounded factuality
• Runs at 1,200 tokens/sec on Cerebras inference hardware
• Model family: Sonar, Sonar Pro, Reasoning Pro, Deep Research
• Matches GPT-4o on user satisfaction benchmarks

AI-native coding model built for software engineering.

• Mixture-of-experts model trained via RL in real development environments
• 4x faster generation than comparable frontier coding models
• Background agents run tasks autonomously while you work
• Cursor Automations trigger agents from GitHub PRs, Slack, Linear, PagerDuty

The leading open-source model family.

• Llama 4 Scout: industry-leading 10M token context window
• Fully open weights; can be self-hosted for complete data control
• Performance competitive with paid frontier models
• Requires powerful hardware for full-scale deployment

Frontier performance at a fraction of the cost.

• 671B parameter MoE model famously trained for under $6M
• R1 variant excels at transparent, step-by-step reasoning
• V3.2-Speciale matches GPT-5 level performance on key benchmarks
• Open-source with competitive benchmark scores; V4 multimodal imminent

The enterprise-safe European choice.

• Strong open-weight models designed for EU AI Act compliance
• Default choice for regulated industries (finance, healthcare, gov)
• Good multilingual support, especially European languages
• Balances capability with privacy and sovereignty requirements

Open-source agentic powerhouse with Agent Swarm.

• 1T total parameters (32B active), MoE architecture with native vision-language
• Agent Swarm coordinates up to 100 specialized sub-agents in parallel
• Kimi Code CLI agent rivals Claude Code and Gemini CLI
• Backed by Alibaba and HongShan; strong global traction

Frontier-class model on a MIT license.

• 744B parameter MoE model (44B active) with 200K context window
• Released under MIT license; trained entirely on Huawei Ascend chips
• 77.8% on SWE-bench Verified; 50.4% on Humanity's Last Exam
• Priced roughly 6x cheaper than comparable proprietary models

The multilingual giant.

• Supports 119 languages with hybrid Mixture-of-Experts architecture
• 0.6B to 235B open-weight; Qwen3-Max (1T+) API-only; Qwen3-Coder (480B) for code
• Competitive with DeepSeek-R1 and OpenAI o1 on reasoning benchmarks
• Qwen3-Coder achieves 69.6% on SWE-Bench Verified, surpassing many frontier models