Grok Models: Full List, Key Features & Use Cases

xAI's Grok model family has grown quickly since its first release in late 2023, moving through frequent point updates, a dedicated "Fast" line for high-volume use, and multi-agent configurations for complex reasoning. Along the way, xAI merged with SpaceX and now operates under the name SpaceXAI, and acquired the coding startup Cursor — a partnership that directly shaped its newest model. This guide lists every major Grok model in order from the newest to the oldest, with a concise overview of each model's specs, key features, and intended use — plus links to try the ones available on Chat Smith.

Released July 8–9, 2026, Grok 4.5 is xAI's newest flagship, built on a new V9 architecture (a mixture-of-experts model reported around 1.5 trillion parameters) and jointly trained with Cursor on real developer-agent data. Musk describes it as "Opus-class, but faster, more token-efficient and lower cost." It ranks #4 of 168 models on the Artificial Analysis Intelligence Index. Context window: 500K tokens. Pricing: $2/1M input, $6/1M output ($0.50 cached input). Configurable reasoning effort: low, medium, or high. Best for: coding and agentic work where token efficiency matters.

Released in beta April 17, 2026 and moved to general availability in early May, Grok 4.3 added native video input, document generation for PDFs, spreadsheets, and slides, improved tool-calling, and a larger 1M-token context window. Pricing: $1.25/1M input, $2.50/1M output, with reasoning effort configurable from none to high. Best for: long-document analysis and agentic workflows needing more context than Grok 4.5 offers.

Announced February 17, 2026 and refined through beta updates like Beta 2 on March 3, Grok 4.20 introduced a four-agent collaboration system — Grok (coordinator), Harper (fact-checking), Benjamin (technical analysis), and Lucas (creative input) — that cross-verify each other before producing a final answer. Its multi-agent configuration supports up to a 2M-token context window. Pricing: $1.25/1M input, $2.50/1M output. Best for: long-document pipelines and complex agent orchestration needing very large context.

Grok 4.1 launched November 17, 2025 as an incremental update improving reasoning, multimodal understanding, personality, and emotional intelligence over Grok 4. Grok 4.1 Fast followed two days later with a 2M-token context window, roughly 3x fewer hallucinations than Grok 4 Fast, and pricing around $0.20/1M input, $0.50/1M output. Best for: high-volume agentic tasks like customer support and research at low cost.

Released fall 2025, Grok 4 Fast is a speed- and cost-optimized version of Grok 4, with a rapid time-to-first-token and a 2M-token context window at roughly $0.20/1M input, $0.50/1M output. Best for: real-time applications, enterprise data extraction, and summarization where low latency matters more than deep reasoning.

Released July 9, 2025, Grok 4 was xAI's flagship single-agent reasoning model, skipping a numbered "Grok 3.5" release entirely. It scored 87.7 on GPQA Diamond and led frontier models on Humanity's Last Exam at the time. Context window: 256K tokens. Pricing: $3/1M input, $15/1M output (double above 128K tokens). Grok 4 Heavy, available via the $300/month SuperGrok Heavy plan, adds a multi-agent configuration for the hardest reasoning tasks. Best for: deep technical reasoning, math, and research.

Released February 2025, Grok 3 introduced DeepSearch, DeeperSearch, a Think reasoning mode, and reinforcement learning at post-training — features that shaped every Grok release since. Context window: 131K tokens. Pricing: roughly $0.30/1M input, $0.50/1M output. Best for: reference only today, though it remains xAI's free-tier model on X.

Released August 2024, Grok 2 was xAI's second-generation model, bringing meaningful gains in reasoning and coding over Grok-1.5 along with an image generation feature built into X. Grok 2 mini offered a faster, cheaper alternative for simpler queries. Best for: historical reference — fully superseded by the Grok 4.x line.

Released April 2024, Grok-1.5 Vision added multimodal understanding to the Grok-1.5 base model, letting it interpret images, diagrams, and screenshots — useful for explaining charts or converting visual information into text. Best for: historical reference — superseded by later multimodal Grok releases.

Released March 2024, Grok-1.5 improved on Grok-1's reasoning and coding performance and extended its context window well beyond the original model. Best for: historical reference only.

Released November 3, 2023, Grok-1 was xAI's first model and the foundation the entire Grok line built on. xAI later open-sourced its weights and architecture under an Apache 2.0 license. Best for: historical and research reference only.

xAI's Grok lineup now spans more than a dozen releases, from the free-tier Grok 3 to the Cursor-trained Grok 4.5 flagship. For teams and individuals who want to use these models without managing a separate xAI account, Chat Smith gives you access to the available Grok models alongside GPT, Claude, Gemini, and DeepSeek — all from a single interface. Browse the full model catalog on Chat Smith to find the right model for your workflow.

1. What is the newest Grok model in 2026?

As of July 2026, the newest Grok model is Grok 4.5, released July 8–9, 2026. It's xAI's flagship model, built on a new V9 architecture and co-trained with Cursor for coding and agentic work.

2. Which Grok models are available on Chat Smith?

Chat Smith currently offers Grok 4.5, Grok 4, and Grok 4 Fast. You can browse the full list on the Chat Smith model page.

3. What's the difference between Grok and Grok Fast models?

Standard Grok models (like Grok 4 or Grok 4.5) prioritize maximum reasoning depth and are tuned for complex, high-stakes tasks. Fast variants trade some reasoning depth for much lower latency and cost, and typically ship with larger context windows, making them better suited to high-volume, real-time applications.