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-5.3: 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.

GPT-5.6 Terra: CI-First Benefit Score 6.0/10, Humics-Neutral. OpenAI's mid-tier large language model balances intelligence and cost between the flagship GPT-5.6 Sol and the cost-optimized GPT-5.6 Luna, offering a 1.05M token context window and configurable reasoning effort at half the price of the flagship.

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.

Llama 4 Scout earns a CI-First Benefit Score of 5.5/10 and a Humics-Neutral badge. Meta's 109B-parameter Mixture of Experts model activates 17B parameters per token, accepts text and images, and supports a 10M-token context window. It suits long-document, multimodal, and self-managed AI work when a qualified human verifies every result.

Qwen3.8 27B: CI-First Benefit Score 6.0/10, Humics-Neutral. Alibaba's 27B Mixture-of-Experts model balances reasoning, multilingual coverage, and local deployment for learners who need a capable model without frontier pricing.

Command A earns a CI-First Benefit Score of 5.3/10 and a Humics-Neutral badge. Cohere's 111B open-weight model supports a 256K-token context, tool use, retrieval-augmented generation, multilingual work, and private deployment. It fits controlled enterprise and academic workflows when a qualified human verifies every important result.

Phi-4 earns a CI-First Benefit Score of 5.3/10 and a Humics-Neutral badge. Microsoft's 14B dense decoder-only Transformer uses a 16K context window, an MIT license, and text-only input and output. It suits verified math, coding, analysis, and tutoring work through local runtimes or Microsoft Foundry, but its factual and independent evaluation results require disciplined checking.

Gemma 3 earns a CI-First Benefit Score of 6.0/10 and a Humics-Neutral badge. Its 1B, 4B, 12B, and 27B variants support a useful range of local and hosted tasks. Larger variants add image-text input and a 128K context window, but every output still needs source checks and human review.

GPT-5.6 Sol: CI-First Benefit Score 6.0/10, Humics-Neutral. OpenAI's flagship reasoning model with a 1M token context window, multi-modal inputs, and configurable reasoning effort. Released July 2026 at $4/M input and $20/M output tokens.

Kimi K3: CI-First Benefit Score 6.5/10, CI-First Strong. A 2.8-trillion-parameter open-source LLM from Moonshot AI with a 1M-token context window, designed for long-horizon coding and agentic knowledge work. Scores high on time savings and output quality but carries elevated Skill Illusion risk for users who delegate coding without learning.

Hyperagent (Airtable AI Agents): CI-First Benefit Score 7.0/10, Humics-Neutral. An enterprise autonomous agent platform that deploys persistent cloud agents for document analysis, web research, and content generation at scale inside Airtable bases. Best for teams already modeling operational data in Airtable. Centaur mode required: the human reviews every agent action.