Grok 4.20 0309 v2

Grok 4.20 0309 v2, as listed on Artificial Analysis, is a Language Model reasoning system from xAI that delivers high-intelligence text and image understanding with a 2M token context window, with pricing available on a paid per-token basis through xAI's first-party API. It targets developers, AI engineers, and enterprises building reasoning-heavy applications such as code generation, scientific analysis, and long-document comprehension.

On Artificial Analysis, Grok 4.20 0309 v2 is benchmarked alongside hundreds of tracked models, where it ranks among xAI's top-tier reasoning offerings competing with systems from OpenAI, Anthropic, Google, DeepSeek, and Alibaba. The model is evaluated on the Artificial Analysis Intelligence Index v4.0, which aggregates 10 demanding benchmarks including GDPval-AA, τ²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, and CritPt. Its key differentiators are a 2M-token context window — substantially larger than the context windows offered by most competing flagship reasoning models — and fast output speed measured in tokens per second after the first streaming chunk is received.

Compared to other reasoning language models in our directory, Grok 4.20 0309 v2 emphasizes the price-quality frontier: Artificial Analysis charts it on the "most attractive quadrant" of the intelligence-vs-price log scale, where price is computed as a 3:1 blend of input and output token costs. At approximately $3.00 per million input tokens and $15.00 per million output tokens, with a discounted cached input rate of $0.75 per million tokens, it is competitively positioned for high-volume reasoning workloads. The model supports both text and image inputs, making it suitable for multimodal reasoning tasks such as chart interpretation, diagram analysis, and visual question answering. For teams already using xAI's ecosystem or building on top of X (formerly Twitter) data integrations, Grok 4.20 0309 v2 offers a tightly integrated alternative to OpenAI's o-series and Anthropic's Claude reasoning lines, with the trade-off being a smaller third-party provider ecosystem and less mature tooling.