The encyclopedia

Amazon's own foundation model family — Nova Micro / Lite / Pro / Premier / Canvas / Reel — available through Amazon Bedrock with full AWS data-residency + compliance guarantees.

ChatGPT is OpenAI's consumer-facing chat product (Free, Plus, Pro, Team, Enterprise tiers), distinct from the developer API and from individual GPT-N model cards.

Claude.ai is Anthropic's consumer + Pro + Team + Enterprise chat product, distinct from the API-level model cards (Opus / Sonnet / Haiku).

Haiku 4.5 is the small, fast model in the Claude 4 family. Released 2025-10-01. Best for high-throughput latency-sensitive tasks.

Claude Opus 4.7 is Anthropic's current flagship as of May 2026. Ships with a new tokenizer that can produce up to 35% more tokens per input vs Opus 4.6 — per-token price unchanged ($5/$25) but effective request cost can rise.

Sonnet 4.6 is Anthropic's mid-tier model balancing capability and cost. Supports a 1M token context window at the standard price.

Command A is Cohere's enterprise flagship — strong on RAG, agentic workflows, multilingual, and image input. Sustains the $2.50/$10 per million tokens pricing held since the Command R+ Aug 2024 line.

DeepSeek R1 is a reasoning-focused, open-weights model released by DeepSeek. Distributed under an MIT-style license; weights mirror on Hugging Face.

Gemini 2.5 Flash is Google's lower-cost sibling in the 2.5 family. 1M token context window.

Gemini 2.5 Pro is Google's reasoning-focused flagship. 1M token context window. Pricing reported across sources ranges $1.00–$1.25/M input — verify against the live Google AI pricing page before publishing.

Tracking file states Gemini API is used to summarize social media data collected by Google BigQuery; only publicly accessible PII from social media is involved, tool uses a Gemini AI engine, and user input is not used to update the LLMs.

The Gemini consumer app exposes Google's Gemini model family with Free, AI Pro, and AI Ultra tiers; distinct from Vertex AI's Gemini API.

Zhipu AI's GLM-4.5 foundation model. Top open-weights performance from a Tsinghua spinoff, available via API and enterprise on-prem deployment.

GPT-5 launched 2025-08-07. Still widely deployed as a workhorse model. Input $1.25/M, output $10/M.

GPT-5.5 is OpenAI's current flagship model, launched after GPT-5.4 (March 2026). Carries a 2x price increase over GPT-5.4 — input went from $2.50/M to $5/M, output from $15/M to $30/M.

Grok 4.3 launched 2026-04-30. 1M token context window. Cached input at $0.20/M. Requests >200K input tokens billed at 2x.

Internal review email states the platform has Azure OpenAI architecture and a three-layer training opt-out; procurement should initiate MNDA execution to access full subprocessor list and SOC 2 report.

Llama 4 Maverick is Meta's open-weight multimodal flagship, released 2025-04-05. Mixture-of-Experts: 17B active parameters across 128 experts, 400B total. 1M token context window. Distributed under Llama 4 Community License (open weights, not OSI-open-source).

Llama 4 Scout is the long-context sibling to Maverick. 17B active across 16 experts, 109B total. Industry-leading 10M token context window. Released 2025-04-05.

Microsoft Copilot is the consumer-facing chat product (web, Edge sidebar, Windows), with companion M365 Copilot bundles for enterprise productivity.

Mistral Large 3 is Mistral AI's flagship as of 2026. Cited pricing varies across third-party trackers ($0.50/$1.50 vs $2/$6 per 1M tokens) — verify against Mistral's live pricing page before publishing.

Moonshot AI's flagship long-context model (1M+ tokens). Strong long-document reasoning; available in Chinese-market enterprise deployments and via API.

Phi-4 is Microsoft Research's 14B parameter dense decoder-only Transformer. Trained on ~16B tokens (~8.3B unique) from synthetic + curated web data. Released 2024-12-12. Fully open-weights under MIT license — usable commercially without restriction.

Qwen 3 is Alibaba Cloud's third-generation open-weights LLM family. The 235B-parameter model is the flagship; smaller distilled variants exist.

Tencent's Hunyuan foundation model family — large LLM, multimodal Hunyuan-DiT image gen, video gen. Powers WeChat/QQ AI features and the Tencent Cloud enterprise stack.