What is Grok 4.5?
Grok 4.5 is a mixture-of-experts reasoning model that xAI trained with Cursor on coding, science, engineering and maths data. It reads text and images, writes text, and works across a 500K-token context with four reasoning effort levels. There are no downloadable weights; you reach it through the xAI API.
It is no longer xAI's first pick. Grok 4.6 and then Grok 4.7 followed at the same $2 / $6 token price, and xAI now points developers to Grok 4.7 for both code and chat. Grok 4.5 stays on the API, with the cheapest cached input of the three.
Is Grok 4.5 still available?
Yes. Grok 4.5 is still on the xAI API at its launch price, but xAI has since released Grok 4.6 and Grok 4.7 and now recommends Grok 4.7 for code and chat.
- Grok 4.6
- Same $2 / $6 price and 500K context, built on Grok 4.5 with longer training; xAI scores it 65.9% on DeepSWE 1.1 against 54% for Grok 4.5.
- Grok 4.7
- xAI's current flagship at the same price, on a new, larger base model, with no limit on text output.
- Cached input
- Grok 4.5 charges $0.30 per million cached tokens; Grok 4.6 and 4.7 charge $0.50.
- Retirement
- xAI has not announced a retirement for Grok 4.5, and its model page and prices are unchanged.
Who is Grok 4.5 best for?
Grok 4.5 is best for agentic coding on a budget, agent loops that lean on cached input, and teams that want fewer output tokens per task.
Agentic coding on a budget
xAI's own charts put it at 64.7% on SWE Bench Pro and 83.3% on Terminal Bench 2.1, at $2 / $6 per million tokens.
Agent loops heavy on cached input
Cached input costs $0.30 per million tokens, less than on Grok 4.6 or 4.7, which adds up when most of each request is repeated context.
Teams watching output tokens
xAI counts about 16,000 output tokens per SWE Bench Pro task on average, 4.2 times fewer than Opus 4.8 at max effort.
High-volume API workloads
Default limits are 150 requests a second and 50 million tokens a minute, with higher limits on request.
What's included with Grok 4.5?
Four reasoning effort levels, function calling and structured output, image input, xAI's server-side tools and priority processing.
- Reasoning effort
- Four levels, low, medium, high and xhigh, with high as the default.
- Function calling and structured output
- It calls your own tools and can return JSON that follows a schema you set.
- Image input
- It reads JPG and PNG images up to 20MiB each alongside text; output is text only.
- Server-side tools
- Web search and code execution run on xAI's side for $5 per 1,000 calls, on top of tokens.
- Priority processing
- Requests can jump the queue for lower latency, billed at twice the standard token rates.
- Spreadsheets and slides
- xAI trained it to build multi-sheet Excel models, PowerPoint diagrams with native shapes, and Word drafts.
How much does Grok 4.5 cost?
Grok 4.5 costs $2 per million input tokens and $6 per million output tokens on the xAI API, with cached input at $0.30, and double rates once a prompt reaches 200K tokens.
- Cached input $0.30 per million tokens
- 500K-token context, text and image input
- Reasoning effort from low to xhigh, high by default
- Cached input $0.60 per million tokens
- The whole request bills at this rate
Prices are in US dollars on the xAI API, which runs on prepaid credits by default, with monthly invoicing through sales; xAI lists no free credits. Grok 4.5 does not take Batch API requests, so there is no batch discount. Server-side tools cost extra, such as $5 per 1,000 web searches or code runs, and priority processing bills at twice the standard rates. Grok 4.6 and Grok 4.7 cost the same per token but charge $0.50 for cached input. Full pricing on docs.x.ai. We checked these prices on .
How do readers rate Grok 4.5?
No one has rated Grok 4.5 yet. Be the first.
Sign in to rate. One rating per account.
Grok 4.5 pros and cons
It is cheap, fast and sparing with tokens, but newer Grok models cost the same and score higher on xAI's own tests.
Pros
- Low price for its class: $2 / $6 per million tokens, and cached input at $0.30 is cheaper than on Grok 4.6 or 4.7.
- Fast: xAI serves it at about 80 tokens a second, and early Hacker News and Reddit users called it one of the quickest models they had used.
- Sparing with tokens: xAI counts about 16,000 output tokens per SWE Bench Pro task, 4.2 times fewer than Opus 4.8 at max effort.
- Users describe its answers as short and direct, with less padding than Claude or GPT models.
Cons
- Superseded at the same price: Grok 4.6 and Grok 4.7 cost the same $2 / $6, and xAI's own table has Grok 4.6 at 65.9% on DeepSWE 1.1 against 54% for Grok 4.5.
- No Batch API, and once a prompt reaches 200K tokens the whole request bills at double rates, $4 / $12.
- Several Hacker News testers rated it level with, or behind, open models such as GLM 5.2, which costs less per token.
- Cursor said an earlier snapshot of its codebase leaked into training, so Grok 4.5's CursorBench results overstate it, and many launch commenters doubted xAI's self-reported scores.
What are the best Grok 4.5 alternatives?
Three tools on Recaply do Grok 4.5's job. GLM-5.2 is open source, so you can run it on your own server.
Grok 4.6PaidChoose it when you want a newer xAI model at the same $2 / $6 price; xAI's tables score it higher, and it runs on the US regional endpoint for US-only inference.
Grok 4.3PaidChoose it when cost and context matter more than coding strength: it costs $1.25 / $2.50, reads 1M tokens and takes batch jobs at 20% off.
GLM-5.2Open SourceChoose it when you want an open-source model with a 1M-token context for less, at $1.40 / $4.40 per million tokens, and can live with text-only input.
Questions about Grok 4.5
What is Grok 4.5?
Grok 4.5 is xAI's closed coding and agent model on the xAI API, built with Cursor for multi-step engineering work and now two releases behind Grok 4.7 at the same price.
Who is Grok 4.5 for?
Grok 4.5 is best for agentic coding on a budget, agent loops that lean on cached input, and teams that want fewer output tokens per task.
Is Grok 4.5 free?
No. Grok 4.5 is a paid tool, from $2.00 / $6.00 per million input / output tokens.
How much does Grok 4.5 cost?
Grok 4.5's paid plans are Prompts under 200K tokens at $2.00 / $6.00 per million input / output tokens and Prompts of 200K tokens or more at $4.00 / $12.00 per million input / output tokens.
Is there an open source alternative to Grok 4.5?
Yes. GLM-5.2 is an open source tool that does the same job, so you can run your own.
What are the best Grok 4.5 alternatives?
The closest alternatives on Recaply are Grok 4.6, Grok 4.3 and GLM-5.2.
Where can I try Grok 4.5?
On its official website, docs.x.ai.
Should I use Grok 4.5 or Grok 4.7?
For most new work, Grok 4.7: xAI recommends it for code and chat, and it costs the same $2 / $6 per million tokens for prompts under 200K tokens. Grok 4.5 still makes sense when cached input dominates your bill, since it charges $0.30 per million cached tokens against $0.50 on Grok 4.7.
Does Grok 4.5 work with the Batch API?
No. Grok 4.5 does not accept Batch API requests, so every call bills at the standard rate. For large jobs that can wait, Grok 4.3 takes batch requests at 20% off and costs less per token to begin with.
Is Grok 4.5 open source?
No. Grok 4.5 is a closed model with no downloadable weights. You call it through the xAI API under the model name grok-4.5, with grok-4.5-latest as an alias, and pay per token from prepaid credits.
Can I use Grok 4.5 in the EU?
Yes. xAI opened Grok 4.5 to EU users in the API console soon after launch. It runs on xAI's us-east-1 and us-west-2 clusters, and it is not on the US regional endpoint, which serves only Grok 4.6 and Grok 4.7.
How fast is Grok 4.5?
xAI says it serves Grok 4.5 at about 80 tokens per second. It also needs fewer tokens than most rivals, about 16,000 output tokens for an average SWE Bench Pro task, so finished answers tend to arrive sooner.
Our verdict
Grok 4.5 is still a cheap, fast coding model, but most new projects should start on Grok 4.7.
Grok 4.7 costs the same $2 / $6 per million tokens and xAI now recommends it for code and chat; Grok 4.5 keeps an edge only on cached input, at $0.30 against $0.50.
- Skip it for
- New projects, batch jobs, and work that needs more than a 500K-token context.
- Check first
- If your bill is mostly cached input, compare real traffic on Grok 4.5 and Grok 4.7 before you move.
Grok 4.7
Grok 4.20
Muse Spark 1.1
Reflection Beam
DeepSeek V4 Pro
Kimi K3.1 Pro