Hermes Agent by Nous Research: CI-First Benefit Score 6.5/10 (CI-First Strong), Humics-Neutral. The open-source, self-improving AI agent platform that University 365 runs its entire 14-department fleet on. Persistent memory, self-created skills, 20+ messaging platform gateway, cron automation, and Kanban task routing. Free under MIT license.

GPT 5.4: CI-First Benefit Score 5.8/10, Humics-Neutral. An earlier-generation OpenAI reasoning model with a 1.1M token context window, chain-of-thought reasoning, and multimodal input. Released March 2026, now positioned between the flagship GPT 5.6 family and the cost-efficient GPT 5.5.

Muse Spark 1.2: CI-First Benefit Score 5.8/10, Humics-Neutral. A proprietary multimodal reasoning model by Meta Superintelligence Labs with a 1M token context window, agentic orchestration, and computer-use capabilities. Released August 5, 2026.

Kimi K2.7 Code: CI-First Benefit Score 5.5/10, Humics-Neutral. Moonshot AI's coding-focused open-weight model with 1T total parameters (32B active), 256K context, and native INT4 quantization. Designed for agentic software engineering workflows with forced thinking mode and MCP tool integration.

MiniMax M3: CI-First Benefit Score 6.1/10, Humics-Neutral. A frontier multimodal MoE model with 1M context window, 428B total parameters, and 23B active parameters, built for coding, agentic workflows, and long-context tasks. Available via API and open-weights.

Gemini 3.6 Flash: CI-First Benefit Score 5.3/10, Humics-Neutral. Google's fast multimodal model in the 3.6 series, released July 21, 2026. Accepts text, image, speech, and video input with high and low effort variants. Performance benchmarks place it between Gemini 3.7 Flash and Claude Sonnet 5. Available on Google AI Studio.

Claude Haiku 4.5: CI-First Benefit Score 6.8/10, Humics-Neutral. Anthropic's latest compact model matches Sonnet 4 on key coding benchmarks at one-fifth the cost. Built for developers who need speed, volume, and 200K-token context without premium pricing.

Gemini 3.1 Pro: CI-First Benefit Score 5.0/10, Humics-Neutral. A Large Language Model, AI Model for academic and professional use.

Gemini Notebook (NotebookLM): CI-First Benefit Score 6.3/10, Humics-Friendly. Google's source-grounded AI research assistant that summarizes documents, generates audio overviews, and answers questions with citations from your own uploaded sources.

Gemini 3.5 Flash-Lite: CI-First Benefit Score 5.5/10, Humics-Neutral. Google's ultra-efficient edge model for high-volume text and image processing with low latency and cost-optimized pricing. Available on Google AI Studio.

Nemotron 3 Ultra: CI-First Benefit Score 5.8/10, Humics-Neutral. NVIDIA's open-weights Mixture of Experts model with 550B total parameters (55B active), 512K-token context window, and competitive API pricing at $0.60/1M input and $3.60/1M output. Available on Ollama, OpenRouter, and build.nvidia.com.

Gemini 3.6 Thinking: CI-First Benefit Score 5.8/10, Humics-Neutral. Google DeepMind's reasoning model with extended thinking, 1M token context, multimodal input, and competitive pricing at $0.75/1M input tokens.

DeepSeek V4 Pro: CI-First Benefit Score 5.5/10, Humics-Neutral. A Large Language Model for academic and professional use.

Claude Fable 5: CI-First Benefit Score 5.2/10, Humics-Neutral. Anthropic's highest-capability model built for long-running agents, with a 1M-token context window, adaptive thinking, and 128K max output, priced at $10/$50 per million input/output tokens.

Mistral Large 3 earns a CI-First Benefit Score of 7.0/10 and a Humics-Neutral badge. Its 675B-parameter sparse architecture, 41B active parameters, 256K context window, multimodal input, and Apache 2.0 weights support demanding research, coding, multilingual, and enterprise workflows. Strong human review remains necessary because fluent output can still contain factual, analytical, and code errors.

DeepSeek V4 Flash earns a CI-First Benefit Score of 6.3/10 and a Humics-Neutral badge. Its 284B-parameter MoE design activates 13B parameters per token, supports a 1M-token context window, and reaches 125.1 tokens per second in independent testing. It suits high-volume reasoning and coding work when you keep human verification in the loop.

Qwen3.8 Max: CI-First Benefit Score 5.5/10, Humics-Neutral. Alibaba's flagship closed-weights LLM with a 1M token context window, multimodal vision capabilities, and competitive intelligence ranking at #7 globally on Artificial Analysis.

GLM-5.2: CI-First Benefit Score 5.8/10, Humics-Neutral. Zhipu AI's open-weights bilingual large language model with a 1M-token context window, 753B total parameters (40B active, Mixture of Experts architecture), and strong coding and reasoning benchmarks under an MIT license.

GLM-5.3: CI-First Benefit Score 5.8/10, Humics-Neutral. A Large Language Model for academic and professional use.

Grok 4.6: CI-First Benefit Score 6.8/10, Humics-Neutral. xAI's frontier model for coding, agentic tasks, and knowledge work with a 500K context window, configurable reasoning, and real-time web search. Ranked #6 on the Artificial Analysis Intelligence Index with a score of 61.

Gemini 2.5 Flash-Lite: CI-First Benefit Score 4.5/10, Humics-Neutral. Google's fastest and most budget-friendly multimodal model for high-volume, low-latency tasks, priced at $0.10 per 1M input tokens with a 1M token context window.

Grok 4.5: CI-First Benefit Score 5.8/10, Humics-Neutral Badge. Grok 4.5 is xAI's reasoning-capable large language model with a 500K token context window, multi-effort reasoning, and multimodal text-plus-image input. It serves as the balanced workhorse between Grok 4.3 and the newer Grok 4.6.

Claude Sonnet 5: CI-First Benefit Score 6.5/10, Humics-Neutral. Anthropic's most agentic Sonnet model, combining near-frontier reasoning with adaptive thinking effort, a 1M token context window, and $2/$10 per million token pricing for coding, agents, and professional workflows.

Gemini 3.7 Flash: CI-First Benefit Score 6.3/10, Humics-Neutral. Google's latest and most capable Flash model, built for complex coding, agentic workflows, and reliable multi-step execution with a 1M token context window.