All LLMs LLMDex tracks

OpenAI's mid-cycle GPT-5 refresh, improved reasoning, tool use, and multimodal grounding over the 2025 launch.

OpenAI's late-2025 standalone reasoning model, an evolution of o3 with deeper chain-of-thought and stronger multimodal reasoning.

OpenAI's unified flagship combining GPT-line breadth with built-in reasoning, replacing both GPT-4o and the o-series for most users.

GPT-5's mid-tier sibling, most of the quality at a fraction of the price, ideal for high-volume production workloads.

OpenAI's smallest GPT-5 variant, built for ultra-low-cost classification, routing, and high-volume inference.

Smaller, faster, cheaper member of OpenAI's reasoning-model family, great latency-cost balance for hard tasks.

OpenAI's flagship reasoning model, set the bar for hard math, GPQA, and agent benchmarks in 2025.

OpenAI's 2025 GPT-4.x refresh, long-context, fast, still widely deployed even after GPT-5.

Smaller, faster reasoning model, popular as the budget thinking-model option throughout 2025.

GPT-4o's small sibling, defined the cheap-mid-tier slot for most of 2024-2025.

OpenAI's first natively-multimodal model, voice, vision, and text in one network.

Anthropic's mid-2026 flagship, ahead on SWE-bench, agent reliability, and writing quality.

Anthropic's mid-tier 4.6 release, the workhorse model behind most production Anthropic deployments.

Anthropic's smallest 4-tier model, fast and cheap with the family's signature tone.

Anthropic's mid-2025 flagship, the model that established Claude's lead on coding agents and SWE-bench.

Mid-2025 mid-tier Claude, the predecessor workhorse to Sonnet 4.6 and still common in production.

The first Claude with an extended-thinking mode, ushered the reasoning-model paradigm into Anthropic's lineup.

Late-2024 small Claude, fast and cheap with surprisingly strong code quality.

Anthropic's late-2024 mid-tier, set the bar on coding, agents, and tool-use through 2025.

Google's late-2025 flagship, set new benchmarks on long-context, vision, and reasoning at competitive pricing.

Google's high-speed, low-cost mid-tier with the same massive context window, popular for high-volume RAG.

Mid-2025 fast tier, set the bar for cost-efficient long-context generation.

Google's mid-2025 flagship, the model that brought Gemini decisively back to parity with the OpenAI and Anthropic frontier.

Early-2025 fast Gemini, first model with full 1M-token context at the Flash price point.

Google's 2B Gemma, built for laptop and phone inference under tight memory budgets.

Google's mid-2024 open-weight 9B, strong quality for its size, friendly license.

Larger Gemma 2, competitive with Llama 70B on some benchmarks at half the size.

Meta's flagship open-weight model, sparse MoE design competitive with closed-frontier flagships.

Meta's mid-tier Llama 4, the practical workhorse for self-hosted deployments.

Meta's small Llama 4, built for on-device and edge inference.

Meta's late-2024 70B refresh, much-improved over 3.1 with better instruction-following and tool-use.

Meta's first open-weight vision-language model at 90B parameters.

Tiny Llama for mobile and edge, runs comfortably on a phone after quantization.

First open-weight model to match GPT-4-class quality on standard benchmarks.

First open-weight reasoning model to match o1, the release that proved RL-from-scratch reasoning training was reproducible.

DeepSeek's flagship 671B-parameter MoE, frontier-level quality at a tiny fraction of frontier prices.

DeepSeek's code-specialized model, strong on a broad set of programming languages and FIM tasks.

Alibaba's flagship open-weight Qwen3, strong on multilingual, code, and math, Apache-2.0 licensed.

Alibaba's mid-size Qwen3, sweet spot for self-hosting at modest hardware budgets.

Open-weight code specialist, frequently the top open option for self-hosted code completion.

The previous-generation Qwen flagship, still widely deployed for stability.

Small Qwen, practical default for laptop and edge inference.

Top open-weight vision-language model, strong on document understanding and chart analysis.

Mistral's code-specialized model, fast inline completion and strong fill-in-the-middle support.

Mistral's mid-tier balanced model, production-ready at competitive pricing.

Mistral's 124B vision-language model, strong on dense-text document tasks.

Mistral's 8B edge model, designed specifically for on-device and on-prem deployment.

Mistral's small-tier API model, fast and cheap for routing and high-volume tasks.

Mistral's 12B multimodal, first vision-capable Apache-2.0 model from the company.

Mistral's flagship API model, strong on code and reasoning, EU-friendly hosting.

12B model co-built with Nvidia, strong small-model multilingual performance.

Mistral's largest open-weight MoE, Apache-2.0, still widely deployed.

xAI's mid-2025 flagship, top scores on Humanity's Last Exam at launch, with native real-time X integration.

xAI's first frontier-tier release, established the company's Colossus-trained model line.

xAI's first widely-available model, free on X for Premium subscribers and competitive with the GPT-4 mid-tier of its era.

Microsoft's 14B model, exceptional quality-per-parameter via curated synthetic training data.

14B Phi-3.5, predecessor to Phi-4 with strong benchmark efficiency for its size.

Cohere's massively multilingual open-weight model, strong on 23 languages.

Smaller Aya Expanse, multilingual on a single-GPU budget.

Cohere's flagship optimized for RAG and tool use in enterprise settings.

Refreshed Command R with improved tool-use, JSON-mode, and Asian-language support.

Cohere's mid-tier RAG-optimized model, affordable and reliable on retrieval workloads.

AI21's hybrid SSM-Transformer with a 256k context window, strong on long-doc tasks.

Smaller hybrid SSM-Transformer model, fast and efficient at long contexts.

Perplexity's premium answer model, deeper search, more sources, longer answers.

Perplexity's flagship answer-engine model with built-in web search grounding.

Nvidia's 340B open-weight model, useful as a synthetic-data generator and benchmark.

Zhipu AI's flagship, strong open-weight Chinese model with broad commercial deployment.

Reka's 21B reasoning model, Apache-2.0 with extended thinking support.

IBM's enterprise-tuned open-weight model, Apache-2.0 with extensive code training.

IBM's smallest Granite, Apache-2.0, edge-friendly, enterprise-supported.

TII's 2024 open-weight refresh, Apache-2.0, multilingual, and competitive at 10B size.

Amazon's mid-tier multimodal, competitive pricing, deep AWS integration.

Amazon's cheap multimodal tier, under-a-cent-per-million-tokens for basic tasks.

Amazon's text-only ultra-cheap tier, best for high-volume routing and classification.

Allen AI's fully-open language model, Apache-2.0, with reproducible training pipeline.

HuggingFace's tiny model line, punches above its weight on a strict on-device budget.

01.AI's API-tier Chinese-leaning model, strong on Chinese benchmarks at competitive pricing.

Allen AI's vision-language model, open everything (weights, data, training code).

Databricks' 132B MoE, a notable 2024 open-weight release tuned for enterprise.