Grok 4.5 (xAI)

Grok 4.5 is a mixture-of-experts large language model from xAI engineered for coding, agentic workflows, and general knowledge work. According to the leak coverage, it is built on a 1.5T-parameter V9 foundation roughly three times the size of the prior v8-small architecture, with supplemental training that incorporated Cursor data and reinforcement learning to sharpen its coding and agent capabilities. The model was reportedly in private beta at SpaceX and Tesla ahead of its broader release, with internal evaluations described by Musk as performing close to or beyond Opus-class competitors.

In benchmark assessments, Grok 4.5 scores 54 on the Artificial Analysis Intelligence Index, placing it fourth overall behind Fable 5, GPT-5.5, and Opus 4.8. Configurable reasoning effort lets developers dial compute to the task, making it well suited for agent pipelines, tool-driven workflows, and developer-facing assistants that benefit from the Cursor-influenced training. Pricing falls in the mid-premium tier, and the substantial context window supports long-running multi-step tasks. A larger V-series run aimed at an August timeframe hints at continued scaling of this lineage.

LLM Gatewayxai/grok-4-5grok

Open provider page Provider docs

Quick Info

Provider: LLM Gateway
Model key: xai/grok-4-5
Release date: Jul 8, 2026
Last updated: Jul 8, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $2.00
Output token cost: $6.00

Limits

Output tokens: 500,000 tokens
Context window: 500,000 tokens

Transparent token rates

Compare Grok 4.5 (xAI) pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Grok 4.5 (xAI)

LLM Gateway

CoverageRelease Notes

DataNorth's news report from July 13, 2026 covers the Grok 4.5 release with concrete technical detail: a 500K-token context window, 80 tokens per second throughput, $2 per million input tokens and $6 per million output tokens, and a text-and-image input / text-output modality. It confirms function calling, structured o The report also supplies benchmark numbers developers can compare against alternatives served via LLM Gateway: DeepSWE 1.0 at 62.0% in xAI's harness (53% on DataCurve's mini-swe-agent run for DeepSWE 1.1), Terminal Bench 2.1 at 83.3%, and SWE Bench Pro at 64.7%, with training on tens of thousands of NVIDIA GB300 GPUs.

LLM Gateway

CoverageBenchmark

TechJack Solutions' breakdown of Grok 4.5 anchors the model's launch on July 8-9, 2026, citing xAI's own documentation at docs.x.ai for a 500,000-token context window and pricing of $2 per million input tokens and $6 per million output tokens for prompts under 200K. The article frames the model as xAI's replacement for For LLM Gateway users evaluating Grok 4.5 directly via the xAI API, the breakdown consolidates the verifiable specs in one place: 1.5-trillion-parameter V9 foundation, $2/$6 pricing tier, 500K context, and a coding-and-agent positioning benchmarked against Claude Opus 4.8. The piece is a third-party review rather than

LLM Gateway

CoverageRelease Notes

Opper AI's changelog lists Grok 4.5 (xAI) among newly added models on its routing platform, sitting alongside Kimi K3 (Moonshot, direct), Gemini 3.6 Flash, and several other model additions in the same release wave. The changelog entry confirms direct, on-platform availability of Grok 4.5 through Opper's OpenAI-compati For LLM Gateway users evaluating Grok 4.5 routing options, the Opper listing is a useful signal of multi-gateway adoption: the same model is being picked up across independent inference providers in the same window, with EU-hosted infrastructure variants already in play for other xAI models on the same platform. No pri