AI MODEL PROFILE

Grok 4.5

Strong published results on agentic coding benchmarks using far fewer tokens than rivals, in a text only package.

Source review Last reviewed August 7, 2026 By xAI

The facts at a glance

ProviderxAI
TypeText reasoning model
Current version or tiergrok-4.5, released July 2026
ModalityText in, text out
AccessProprietary
Context or limits500,000 token context, knowledge cutoff 1 February 2026
Free optionYes, for a limited time in Grok Build and Cursor, plus grok.com, X, iOS, and Android
Paid priceUSD 2.00 per million input and USD 6.00 per million output

Pricing last checked August 7, 2026. Prices change and vary by region. Check the official pricing page ↗

Assessment

What it does well

  • xAI reports 83.3 percent on Terminal Bench 2.1 and 62.0 percent on DeepSWE 1.0
  • Reported to use roughly 4.2 times fewer tokens than comparable models, which changes the real cost per task rather than just the headline rate
  • Cheapest output pricing of the major proprietary frontier models

Where it falls short

  • Text only. Image, video, and voice require separate Grok Imagine and Grok Voice APIs
  • The 500,000 token context is half or less than the OpenAI, Anthropic, Google, and leading Chinese flagships
  • xAI's own pages give three different release dates for this model, which does not inspire confidence in its documentation

Privacy and data control

xAI states it may use prompts, searches, and responses to train its models, with an opt out under Settings, Data Controls, Improve the model, and a Private Chat mode. Its privacy policy explicitly tells users not to include personal information in prompts.

The longer read

Token efficiency is a real cost lever

Comparing models on price per million tokens assumes they use a similar number of tokens to do the same job. xAI reports Grok 4.5 uses roughly a quarter of the tokens of comparable models on its benchmarks. If that holds for your workload, the effective cost gap is much wider than the published rates suggest. It is worth measuring on your own tasks rather than taking either figure on trust.

How it performs by outcome

We assess products against a job, not a leaderboard. These are the outcomes this entry is compared under.

Worth comparing against

DeepSeek-V4-Flash Teams that need to run a capable model on their own infrastructure Qwen3.8-Max Multi day autonomous coding projects Kimi K3 Research teams that need frontier scale weights they can inspect Mistral Medium 3.5 Coding agents running on a small GPU cluster

How this profile was produced

Source review. Every fact here is drawn from the provider documentation linked below and checked on the date shown. We have not yet run this entry through a controlled test, and no scoring is implied.

Read the full review methodology →

Primary sources