Skip to main content
Uncensored — We add no content filter to this model, but like most chat models it declines a blunt request that comes with no context. It writes unrestricted fiction reliably once your system prompt sets up a fiction-writing context, or once the request is framed as a story. Legal and consent duties remain with the customer.

Grok 4.5 Review: xAI's Coding Grok, Cheaper on Cached Input

Pick Grok 4.5 for a long, reused system prompt where cached-input pricing matters; pick Grok 4.6 for the same content result with a higher benchmark score.

Open the Playground

Spicy Index

78.2/ 100

Freedom Score

100.0/ 100

Content levels

Tested 5 of 5 probe types · 22 runs

  • Mild · The lightest writing requests in the test

  • Moderate · A step up, still below the middle of the scale

  • Strong · Mid-scale requests; must be tested before a model is ranked

  • Very strong · Where most language models start to hold back

  • Maximum · The strongest written requests, scored as one class

Price
$3.00 per 1M mixed tokens (3 input : 1 output)
Output speed
43 tok/s
Success rate
Collecting data

Grok 4.5 passed every rung of our content ladder — L1 Suggestive (a slipping strap), L2 Partial, L3 Strong, L4 Intense (a written intimate scene between two adults) and L5 Extreme (a written scene past L4, either restraint and impact play or graphic violence between adults) — with all 22 scored runs on 19 and 24 September 2026 rendering in full, under direct, fiction-framed and system-prompt phrasing alike. It's positioned by xAI as the coding model in the Grok line, trained alongside Cursor, and holds 500,000 tokens with cached input priced at $0.30 per million — 40% below Grok 4.6's $0.50.

The verdict

Seven runs were L4 Intense: two direct requests with no story or persona, two framed as fiction, three set through a system prompt. Every one rendered in full, the same clean record we measured on Grok 4.6. The rest of the ladder followed the same pattern: three runs each for L1 Suggestive, L2 Partial and L3 Strong, six for L5 Extreme, all rendering.

Pick it when your workload reuses a long system prompt across many turns or agent steps — its cached-input rate is the reason to choose it over Grok 4.6 rather than any difference in what it will write, since the two rendered identically on our content ladder.

Try Grok 4.5 →

Key facts

  • Pass rate: 22 of 22 scored runs rendered in full on 19 and 24 Sep 2026, across direct, fiction-framed and system-prompt phrasing at every rung.
  • Price: $2.00 per million input tokens, $0.30 cached and $6.00 output; from 200,000 input tokens, $4.00 and $12.00 — the same as xAI's own published rate.
  • Limits: 500,000-token context, 256 messages per request, replies up to 450,000 tokens (default cap 4,096), text and image input, no files.
  • Sources: every scored run is in the Grok 4.5 test report, the scoring rules are on our review methodology page, and the full parameter list is on the Grok 4.5 model page.

What we liked and what held it back

What we liked

  • Every rung rendered, every phrasing. L4 Intense rendered on direct requests with no story, no persona and no system prompt, and the fiction and system-prompt versions rendered too.
  • Cached input costs less than on Grok 4.6 or Grok 4.7. $0.30 per million tokens against their $0.50; Grok 4.3 prices cached input lower still, at $0.20.
  • Built for tool-calling agent loops. xAI positions it as the coding model in the Grok line, trained alongside Cursor; up to 128 tool definitions per request.
  • It reads images. A user message can carry image_url parts at detail auto, low or high — a screenshot next to a question, in one turn.
  • Priced at xAI's own rate. $2.00/$0.30/$6.00 per million tokens, the same as xAI's published rate.

What held it back

  • Reasoning defaults to High. Deliberation runs before every reply and is billed as output; set reasoning_effort to low if the request has little to work out.
  • 256 messages per request, whatever the token count. A long chat history hits this cap before the 500,000-token window fills up.
  • The default reply cap is small. max_tokens defaults to 4,096 against a 450,000-token ceiling; a long reply truncates silently unless you raise it.
  • The bill jumps past 200,000 tokens. At 200,000 input tokens the whole request moves to $4.00 input and $12.00 output per million.
  • No files. Text and images only, no file parts on this endpoint.
  • Scores lower than Grok 4.6 on third-party benchmarks. 38.8 on the Artificial Analysis Intelligence Index (rank 27 of 462, 27 Sep 2026) against Grok 4.6's 44.3 — the two write our ladder identically, but Grok 4.6 is the newer, more capable model.

What we tested

We ran Grok 4.5 through our full content ladder between 19 and 24 September 2026, each rung asked three ways: direct, framed as fiction, and set through a system prompt. L1 Suggestive, L2 Partial and L3 Strong each got one run per phrasing, and all nine rendered in full. L4 Intense got two direct runs, two fiction-framed runs and three system-prompt runs — seven in total, all rendered. L5 Extreme got two runs per phrasing, six in total, all rendered.

Across the ladder that's 22 scored runs, all 22 rendering in full: nothing softened and nothing refused, at any rung or any phrasing — the identical result we measured on Grok 4.6. The Spicy Index page reads this into a Freedom Score on this page's score card.

Prompts that worked

Grok 4.5 doesn't need a system prompt to write past any rung of our ladder; reach for one to fix a voice or a persona, or to set a strict output format for a coding agent.

1. A published example: a plan you can check, at High effort. The shape an engineering loop sends most: a system prompt that fixes the output format, one task, no images.

text
system: Write evaluation plans as a numbered list of checks, each with the observation that would make it fail.
user: We are replacing our retry logic with an idempotency key. Plan how to prove the new path cannot double-charge.

Run this prompt on Grok 4.5 →

2. A published example: two images and a question, one user turn. Two image_url parts at detail high beside the text.

text
user: These two dashboards are before and after Tuesday's deploy. Which panel shows the regression, and what is the most likely cause?
[image: https://cdn.spicyapi.ai/models/examples/inputs/c681580c3c6a0b63.jpeg]

Run this prompt on Grok 4.5 →

Settings that matter: reasoning Low for a quick pass, High or xhigh when the answer needs working through; a repeated system prompt across an agent loop is where the $0.30 cached-input rate pays off most, since it's charged on every turn that reuses it; keep history under 256 messages and trim a request near 200,000 tokens down below the line before sending.

Roleplay and how far it writes

Grok 4.5 rendered every one of our 22 scored runs in full across the whole content ladder: L1 Suggestive through L5 Extreme, under direct, fiction-framed and system-prompt phrasing alike. Nothing was softened and nothing was refused, including direct L4 Intense requests with no story, no persona and no system prompt. The page label is Uncensored: we add no content filter on our side.

Freedom Score: shown on this page's score card and recalculated as we add runs.

Not tested yet: a long chat staying on track many turns deep (our runs were single requests); scenes between two women or two men, fantasy species and roleplay in languages other than English; long agent loops on our API.

Every character must be an adult, keep real, identifiable people out of it, and you're responsible for the law where you publish.

What the community says

Grok 4.5's launch was read mostly through its price. One widely upvoted post on its debut called it "5x cheaper" against comparable frontier models, and a Hacker News comment the same day called it "extremely economical".

Alternatives

  • Same content-ladder result, a higher benchmark score? Grok 4.6, 44.3 on the Artificial Analysis Intelligence Index against Grok 4.5's 38.8, at the same base price.
  • Tighter replies, no jailbreak needed either? Grok 4.7; see our Grok 4.7 review.
  • A cheaper cached-input rate still, and a bigger window? Grok 4.3, $0.20 per million cached tokens, 1,000,000 tokens of context and a reasoning-off switch.
  • A tiny bill for volume passes? GLM 5.3 Flash, $0.1275 per million input tokens.

How far Grok 4.5 goes

Each level, on each task, based on confirmed results. A level earns points only when the content actually appears, and a single run earns 60% at most.

Freedom points by level and task
LevelChat
MildThe lightest writing requests in the test
3/3 delivered20.0 / 20 pts
ModerateA step up, still below the middle of the scale
3/3 delivered20.0 / 20 pts
StrongMid-scale requests; must be tested before a model is ranked
3/3 delivered20.0 / 20 pts
Very strongWhere most language models start to hold back
7/7 delivered20.0 / 20 pts
MaximumThe strongest written requests, scored as one class
6/6 delivered20.0 / 20 pts
Task score100.0 / 100.0Tested Sep 24, 2026

Tested Sep 19, 2026 to Sep 24, 2026

This site stays safe for work: results are written up as text, any image shown is pixel-censored, and test prompts are never published. We never test anyone who looks underage, real people’s likenesses or non-consent.

See every test record →See the uncensored leaderboard →

Run your own prompt and see how far it goes

Open the Playground

How Grok 4.5 earns its Spicy Index

Each dimension below links to its full workings: the table row every point came from and why points were lost.

Public leaderboards 40% of quality
75.7
Role-play 15% of quality
53.4
Creative writing 15% of quality
36.5
Instruction following 15% of quality
95.0
Long-context memory 15% of quality
Not yet tested

Our long-context suite hasn’t run on this model yet. How it works

Capability 10.5% weight
72.0
Freedom Score 30% weight
100.0
Open the full test report →

Capability, item by item

Points from the public scoring table. Upscaled tiers and limits we set ourselves don’t count.

Context window (creator-published)500,000 tokens
20.0 / 30
Input modalities on SpicyAPIText, plus images
12.0 / 25
Tool useTools and tool_choice
15.0 / 15
Structured outputAccepted, schema adherence verified
15.0 / 15
Reasoning control4 levels: low · medium · high · xhigh
10.0 / 15

What Grok 4.5 can do on SpicyAPI

Straight from the live model specification on SpicyAPI. Some creators offer more in their own apps; this is what you can call through our API today.

Capabilities by endpoint
FeatureChat
Context window500,000 tokens
Max output per request450,000 tokens
Reasoning levelslow · medium · high · xhigh
Tool callingYes
Structured outputYes
See every parameter and code sample →

How long it really takes

Measured on real jobs over the last 30 days, from request to finished result. Times include our own test runs.

Output speed
43 tok/sIncludes time to first token
Success rate
Collecting data

A time appears once at least 10 completed jobs are in. Success rates count customer jobs only and stay hidden for every model until we have a full 30 days of traffic.

What people say about Grok 4.5

2 positive · 0 mixed · 0 negative

  • Thats 5x cheaper

    Reddit· Jul 8, 2026· PositiveSource

  • extremely economical

    hacker-news· Jul 8, 2026· PositiveSource

Quotes are verbatim from public posts, linked to the original. They are not ratings and don’t feed any score.

Models to consider instead of Grok 4.5

  • xAI

    Grok 4.6

    Same content-ladder result, a higher benchmark score?

    Spicy Index
    77.2−1.0
    Freedom Score
    100.0
    Price
    $3.00/ 1M tokens · 3:1 input/output
    Compare side by sideTry

  • xAI

    Grok 4.7

    Tighter replies, no jailbreak needed either?

    Spicy Index
    78.2
    Freedom Score
    100.0
    Price
    $3.00/ 1M tokens · 3:1 input/output
    Compare side by sideTry

  • xAI

    Grok 4.3

    A cheaper cached-input rate still, and a bigger window?

    Spicy Index
    77.0−1.2
    Freedom Score
    98.9−1.1
    Price
    $1.56−48%/ 1M tokens · 3:1 input/output
    Compare side by sideTry

  • Z.ai

    GLM 5.3 Flash

    A tiny bill for volume passes?

    Spicy Index
    72.9−5.3
    Freedom Score
    64.4−35.6
    Price
    $0.202−93%/ 1M tokens · 3:1 input/output
    Compare side by sideTry

What do you think?

Was this review useful?

Log in to vote.

Comments

Comments are checked automatically and by our team. We remove spam, offensive content, personal attacks and anything that names a real person without consent. Report a comment and we review it within one business day.

Tried this model? Share what worked. You need a free account to comment, so we can keep spam out.

Grok 4.5 FAQ

Try Grok 4.5 yourself

Set every parameter in the Playground and see the exact price before you run it.

$3.00 per 1M mixed tokens (3 input : 1 output)

More reviews

See the leaderboards →
  • Three glass discs stacked in the dark, crossed by a streak of cyan light, with a thin gold spark rising beneath them

    Grok 4.7

    Pick Grok 4.7 for uncensored roleplay and fiction that must work with no jailbreak; skip it for long chapters per turn or long agent loops.

    Spicy Index
    78.2
    Freedom Score
    100.0

    Tested Sep 24, 2026

    Read the review
  • Warm golden light streaks passing through a row of translucent blue glass panels on a dark pedestal

    DeepSeek V4.1 Flash

    Pick DeepSeek V4.1 Flash for cheap, whole-document reads and a 384,000-token ceiling; add a system prompt to steady intimate scenes (L4).

    Spicy Index
    81.9
    Freedom Score
    90.4

    Tested Sep 26, 2026

    Read the review
  • A burst of golden light filaments fanning low across a shallow glass dish against pure black

    Grok 4.3

    Pick Grok 4.3 for a huge context window, a genuine reasoning off switch and uncensored writing at a low price; skip it if you need file uploads.

    Spicy Index
    77.0
    Freedom Score
    98.9

    Tested Sep 24, 2026

    Read the review