xAI (SpaceXAI) — Released July 8, 2026
20 free Grok 4.5 prompts for coding, agentic tasks, research, and knowledge work — with configurable reasoning effort. Copy directly into Cursor or the xAI API. No signup.
The Grok 4.5 prompt generator on this page gives you 20 copy-paste prompts designed to get the most out of xAI's newest AI model. Grok 4.5 (model ID: grok-4-5) launched July 8, 2026, built on xAI's V9 architecture — a 1.5 trillion parameter Mixture-of-Experts foundation co-trained with Cursor on real developer coding sessions. The result is an Opus-class reasoning model optimised for coding, agentic workflows, and knowledge-intensive work, priced at $2/$6 per million input/output tokens.
The prompts on this page are organised by use case — coding, agentic tasks, research, writing, and knowledge work — and include a recommended reasoning effort level for each. Grok 4.5 supports three configurable effort levels (low, medium, high), and specifying the right one in your prompt is the single biggest quality lever. Paste any prompt into the xAI API, Grok Build, or Cursor and fill in the bracketed placeholders with your context.
Simple rewrites, email tone fixes, formatting, quick lookups, and one-sentence answers. No heavy reasoning needed — use this for speed.
Writing, standard code review, SQL optimisation, data summaries, competitive analysis, API documentation. The right default for most tasks.
Complex debugging, research synthesis, multi-step agentic tasks, hard STEM problems, architectural decisions, and PRDs. Use when output quality matters most.
Each prompt shows the recommended reasoning effort level. Works with xAI API, Cursor, Grok Build, and OpenRouter.
Review this code thoroughly. For every bug or issue found, provide: (1) the exact line and what is wrong, (2) why it will fail (include the specific input or condition that triggers it), (3) the corrected code snippet, (4) a one-line note on how to prevent this class of bug. Code: ``` [Paste your code here] ``` Language/framework: [e.g. TypeScript / React 19 / Node 22]
Execute the following multi-step task. Before each step, state: (a) what you are about to do, (b) what information you need, (c) what decision you are making. After completing all steps, write a one-paragraph summary with a clear next action. Task: [describe your goal, e.g. 'Analyse this dataset, find the top 3 anomalies, and write a brief for the data team'] Context and data: [Paste your context, documents, or data here]
Build a production-ready implementation of this feature: Feature: [describe the feature] Stack: [e.g. Next.js 15 + TypeScript + PostgreSQL + Prisma] Deliver: 1. Database schema changes with migration SQL 2. API route(s) with input validation and error handling 3. React component with loading, error, and empty states 4. Unit test stubs for every new function 5. A comment on any non-obvious architectural decision Existing patterns: [paste a sample file or describe conventions]
Solve this step by step, showing every deduction before stating the answer. Do not skip any elimination step. [Paste your logic puzzle or constraint-satisfaction problem here] After your deductions, state the answer clearly and explain which single clue was the decisive one.
Synthesise these three sources on [topic] into a structured brief: 1. CONSENSUS — what all sources agree on 2. CONTRADICTIONS — where they disagree and why the disagreement exists 3. GAPS — what none of them address 4. TAKEAWAY — one-paragraph practical conclusion for a non-specialist Sources: [Source A text or URL] [Source B text or URL] [Source C text or URL]
Write an Architecture Decision Record (ADR) for the following technical choice: Decision: [e.g. 'Use PostgreSQL instead of MongoDB for the user-activity store'] Context: [describe the system, scale, and team constraints] Options considered: [list 2–4 alternatives] The ADR must include: 1. STATUS — proposed / accepted / deprecated 2. CONTEXT — why this decision needs to be made now 3. OPTIONS — pros and cons of each in a table 4. DECISION — which option and why 5. CONSEQUENCES — positive and negative effects 6. REVISIT TRIGGER — what would make us reconsider this decision
Solve this STEM problem. Show every step of your working. After each step, state the principle you are applying (e.g. conservation of energy, chain rule, Bayes' theorem). Flag any assumption you make. Problem: [Paste your maths, physics, chemistry, or engineering problem here] At the end, sanity-check your answer using a different method or by substituting the result back into the original equation.
Write a strategic competitive analysis for the following market: Our product: [one sentence description] Target customer: [describe] Competitors to analyse: [list 4–6 names] For each competitor: - Primary value proposition (one sentence) - Estimated customer profile - One advantage they have over us - One advantage we have over them - Apparent pricing model Close with: the unclaimed positioning in this market — what gap none of these players owns.
Analyse this SQL query for performance and rewrite it to be faster. Explain every change. Database: [PostgreSQL / MySQL / SQLite] Table sizes: [e.g. orders: 15M rows, users: 900K rows] Current query: ```sql [Paste your query] ``` Deliver: 1. The optimised query 2. A bullet list of each change and why it helps 3. Indexes to add (with the exact CREATE INDEX statement)
Diagnose this bug. Walk through every plausible cause systematically before settling on the most likely one. Error: ``` [Paste the error message and stack trace] ``` Relevant code: ``` [Paste the code] ``` What I have tried: [list attempts] Environment: [OS, language/runtime version, key packages] Deliver: 1. Hypothesis ranking (most → least likely, with reasoning) 2. The fix for the top hypothesis with specific line changes 3. Diagnostic steps if the top fix does not work 4. How to prevent this class of bug in future
Write a complete Product Requirements Document for this feature: Feature: [name] Problem it solves: [one sentence] Target user: [describe] Success metric: [how will we know it worked?] Include: 1. PROBLEM STATEMENT — 2 paragraphs 2. PROPOSED SOLUTION — what we are building (not how) 3. USER STORIES — 5 in 'As a [user], I want [goal], so that [reason]' format 4. ACCEPTANCE CRITERIA — testable conditions for each story 5. OUT OF SCOPE — 3 things not in v1 6. OPEN QUESTIONS — 3 unresolved decisions with owners 7. TIMELINE — phases in sequence (no dates)
Analyse this data and explain what it means without jargon. Data: [paste your CSV, table, or key numbers] Answer: 1. The single most important trend 2. The most surprising or counterintuitive finding 3. What the data does NOT tell you that you might wrongly assume 4. What one additional data point would most change the interpretation 5. A two-sentence summary for someone who will not read the full analysis
Rewrite this email to be professional but warm — direct without being cold, confident without being arrogant. Keep the core message identical. Do not add filler phrases like 'I hope this email finds you well.' Flag any sentence that could be read as passive-aggressive. [Paste your email draft]
Generate a complete test suite for this function using [Jest / Vitest / pytest / your framework]. Cover: happy path, edge cases, invalid inputs, boundary conditions, and any async behaviour. Function: ``` [Paste your function] ``` For each test: - Name it descriptively (scenario + expected result) - Use the AAA pattern (Arrange / Act / Assert) - Add a one-line comment explaining why this case matters
Write a 1,500-word article on the following topic: Topic: [your topic] Primary keyword (must appear in H1, intro, and at least twice in body): [keyword] Audience: [describe who will read this and their expertise level] Tone: [conversational / authoritative / technical] Structure: - H1 with the primary keyword - 120-word intro that states the reader's problem and promises a specific takeaway - 4–5 H2 sections with concrete examples - 100-word conclusion with one clear action step No filler phrases, no generic advice, no conclusions that merely restate the intro.
Write complete API documentation for this endpoint in OpenAPI 3.1 YAML, then write a human-readable markdown reference section. Endpoint: - Method: [GET/POST/etc] - Path: [/your/endpoint] - Auth: [bearer token / API key / none] - Request body: [describe fields] - Success response: [describe fields] - Error codes: [list them] The documentation must be clear enough for a developer who has never used this API to succeed on the first request.
Help me make this decision using a structured framework: Decision: [describe it, e.g. 'Build vs buy this analytics feature?'] Constraints: [time, budget, team size] Stakeholders: [who is affected] Options under consideration: [list them] Deliver: 1. A weighted decision matrix with the 5 most important criteria 2. Your recommendation with clear rationale 3. The strongest argument AGAINST your recommendation 4. What new information would change your recommendation
I have an interview for [job title] at [company type, e.g. 'Series B startup' or 'Fortune 500']. My background: [2–3 sentence summary] The role focuses on: [key responsibilities] Generate: 1. The 8 most likely interview questions for this specific role 2. A strong answer framework for each (key points, not a full script) 3. The 3 questions I should ask the interviewer 4. The one weakness to avoid mentioning given this role and company
Write a production-quality system prompt for a Grok 4.5 deployment: Use case: [describe the assistant, e.g. 'technical support agent for a SaaS developer tool'] Tone: [formal / casual / technical / empathetic] Out-of-scope handling: [what should it refuse or redirect?] Knowledge boundaries: [what does it know and not know?] Output format: [how should responses be structured?] The system prompt must: - Define persona and role in the first 3 sentences - Include explicit fallback instructions when the model doesn't know the answer - Add a reasoning effort instruction appropriate to the task - Be under 800 tokens Also write 3 test prompts to verify the system prompt is working.
Audit this codebase for technical debt. For each issue found, provide: (1) file and line reference, (2) what the problem is, (3) risk level (low/medium/high), (4) estimated effort to fix (hours), (5) recommended fix. Present results as a prioritised backlog sorted by risk × effort. Code to audit: ``` [Paste your code or describe the codebase structure] ``` Focus areas: [e.g. security, performance, maintainability, test coverage]
How Grok 4.5 compares to the top text/reasoning AI models across the dimensions that matter for coding, research, and knowledge work.
| Model | Best For | Reasoning | Context | Price (input/output) |
|---|---|---|---|---|
| Grok 4.5 ✓ | Coding, agentic tasks, STEM, knowledge work | Low / Medium / High | 128K tokens | $2 / $6 per MTok |
| Claude Sonnet 5 | Everyday power use, writing, agentic tasks | Low / Medium / High / xhigh | 1M tokens | $2 / $10 per MTok |
| Claude Fable 5 | Hardest frontier tasks, full codebase audits | Adaptive (always on) | 1M tokens | $10 / $50 per MTok |
| GPT-5.6 | General use, image understanding, plugin ecosystem | Standard | 128K tokens | $5 / $20 per MTok |
| Gemini 4 | Multimodal tasks, Google ecosystem integration | Thinking mode (on/off) | 1M tokens | $7 / $28 per MTok |
| DeepSeek V4 | Open-source research, self-hosted deployments | Extended reasoning | 1M tokens | MIT license (free) |
grok-4-5 explicitlyGrok 4.5 is xAI's (SpaceXAI's) newest AI model, released July 8, 2026. It runs on the V9 architecture — a 1.5 trillion parameter Mixture-of-Experts foundation, roughly three times the size of the V8-small architecture behind Grok 4.3. The model was co-trained with Cursor on real developer usage data, making it exceptionally strong at coding and agentic tasks. Elon Musk described it as 'an Opus-class model, but faster, more token-efficient, and lower cost'. It is priced at $2 per million input tokens and $6 per million output tokens.
Grok 4.5 introduces configurable reasoning effort — low, medium, or high — that controls how deeply the model thinks before responding. Low effort gives fast, lightweight responses suited to simple lookups and rewrites. Medium is balanced and works well for writing, analysis, and standard coding tasks. High effort engages the model's full reasoning depth — best for complex debugging, research synthesis, multi-step agentic tasks, and hard STEM problems. Setting the right level makes a measurable difference in both output quality and cost.
Grok 4.5 is a major architectural upgrade from Grok 4.3. The key changes are: the V9 foundation (1.5T parameters, MoE vs. 4.3's smaller V8-small), co-training with Cursor on real developer coding sessions (dramatically improves code quality and developer ergonomics), configurable reasoning effort (4.3 had no explicit reasoning modes), improved STEM and research synthesis performance, and significantly lower cost at $2/$6 per MTok. Grok 4.5 is also the new default model in Grok Build and available natively inside Cursor.
Grok 4.5 is available through multiple channels: the xAI API (direct access, model ID: grok-4-5), Grok Build (xAI's developer platform, where it is the default model), Cursor (built-in for coding workflows), and third-party model gateways including OpenRouter, Vercel AI SDK, Cloudflare Workers AI, Snowflake Arctic, and Databricks Mosaic AI. Note: Grok 4.5 is not yet available in the EU as of July 2026 — EU availability is expected mid-July 2026.
Grok 4.5 is purpose-built for coding, agentic workflows, and knowledge-intensive work. It excels at: full-stack feature development, codebase audits, complex debugging and root cause analysis, multi-step agentic tasks (plan-then-execute), STEM problem solving, research synthesis, and long-document knowledge work. It also performs well at writing, analysis, and decision frameworks. For pure image or video generation, Grok Imagine and Grok Imagine Video are the xAI tools to use — Grok 4.5 is a text-only model.
All 20 prompts on this page are free to copy and use directly — no signup needed on randomprompts.org. Grok 4.5 itself is available via the xAI API at $2/$6 per million tokens (pay-as-you-go) and is built into Cursor (where it is included in Cursor Pro subscriptions). If you have an existing Cursor subscription, you can use Grok 4.5 at no additional cost by selecting it as your model. Copy any prompt from this page and paste it into your Grok 4.5 session, filling in the bracketed placeholders with your specific context.
Anthropic's new default model — 20 reasoning prompts
Anthropic's most capable frontier model — 20 prompts
OpenAI's latest model — 20 copy-paste prompts
Google's flagship multimodal AI — 20 prompts
Microsoft's first in-house reasoning model — 20 prompts
xAI's image generator — 20 AI art prompts