// head_to_head

GPT-5.6 Luna vs SpaceXAI: Grok 4.6

Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.

OpenRouter + LiveBenchAll comparisonsFull leaderboard

SpaceXAI: Grok 4.6 scores higher, GPT-5.6 Luna costs less — it depends on your workload.

SpaceXAI: Grok 4.6 is ahead by 4.5 points overall, and GPT-5.6 Luna lists 6.7× cheaper per blended million tokens. Whether 4.5 points is worth that depends on how much a wrong answer costs you. GPT-5.6 Luna also leads on measured cost per point of capability, at $0.0911 per point.

openai

GPT-5.6 Luna

Blended / 1M
$0.450
Context
1.1M
Released
Jul 9, 2026
Overall score
73.6
reasoningtool callingfile inputimage inputprompt caching

x-ai

SpaceXAI: Grok 4.6

Blended / 1M
$3.00
Context
500K
Released
Aug 12, 2026
Overall score
78.0
reasoningtool callingimage inputfile inputprompt caching

Specs and pricing

MetricGPT-5.6 LunaSpaceXAI: Grok 4.6
LiveBench overall

Mean of the seven LiveBench category scores, 0–100. Higher is better.

73.678.0win
Cost per point

Measured benchmark spend divided by overall score — dollars per point of capability.

$0.0911win$0.1181
Blended price / 1M

3:1 input:output mix, the usual shape of production traffic.

$0.450win$3.00
Input price / 1M$0.200win$2.00
Output price / 1M$1.20win$6.00
Cached input / 1M

Price of an input token served from the prompt cache, where the provider publishes one.

$0.020win$0.500
Context window1.1Mwin500K
Max output tokens128K

Benchmarks by category

An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — GPT-5.6 Luna on top, SpaceXAI: Grok 4.6 below, both out of 100.

Agentic coding
48.4
57.0
Coding
82.9
76.8
Reasoning
85.6
90.5
Mathematics
87.2
92.6
Data analysis
78.0
73.9
Language
72.6
83.7
Instruction following
60.1
71.9

What each one costs to run

Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.

WorkloadGPT-5.6 LunaSpaceXAI: Grok 4.6
Support chatbot

1.2K in / 400 out × 200K requests

$131.04/mo$852.00/mo
RAG assistant

8K in / 600 out × 100K requests

$160.00/mo$1360.00/mo
Coding agent

40K in / 4K out × 20K requests

$155.20/mo$1240.00/mo
Document extraction

20K in / 1.5K out × 50K requests

$281.00/mo$2375.00/mo
Bulk classification

500 in / 20 out × 5M requests

$530.00/mo$4850.00/mo
Run these two through the cost calculator

Which should you pick?

You are running this at volume

GPT-5.6 Luna

Lowest measured cost per point of capability at $0.0911 per point — the gap compounds with every request.

Quality matters more than the bill

SpaceXAI: Grok 4.6

Highest overall LiveBench score of the two at 78.0.

The workload is coding or agentic work

SpaceXAI: Grok 4.6

Leads on agentic coding — 57.0 against 48.4.

You need to fit large documents in one call

GPT-5.6 Luna

Wider context window — 1.1M against 500K.

GPT-5.6 Luna vs SpaceXAI: Grok 4.6 FAQ

Which is better, GPT-5.6 Luna or SpaceXAI: Grok 4.6?

SpaceXAI: Grok 4.6 scores higher, GPT-5.6 Luna costs less — it depends on your workload. SpaceXAI: Grok 4.6 is ahead by 4.5 points overall, and GPT-5.6 Luna lists 6.7× cheaper per blended million tokens. Whether 4.5 points is worth that depends on how much a wrong answer costs you. GPT-5.6 Luna also leads on measured cost per point of capability, at $0.0911 per point.

Is GPT-5.6 Luna cheaper than SpaceXAI: Grok 4.6?

GPT-5.6 Luna is cheaper. On a 3:1 input:output blend, GPT-5.6 Luna lists at $0.450 per million tokens and SpaceXAI: Grok 4.6 at $3.00 — GPT-5.6 Luna is 6.7× cheaper. Input and output are priced separately — GPT-5.6 Luna charges $0.200 in and $1.20 out, SpaceXAI: Grok 4.6 charges $2.00 and $6.00 — so the model that looks cheaper flips depending on how output-heavy your workload is.

GPT-5.6 Luna vs SpaceXAI: Grok 4.6: which scores higher on benchmarks?

GPT-5.6 Luna scores 73.6 and SpaceXAI: Grok 4.6 scores 78.0 overall on LiveBench, the mean of its seven categories. That is a 4.5-point lead for SpaceXAI: Grok 4.6. Category scores differ from the overall figure — a model can lead on reasoning and trail on coding, which the per-category table above breaks out.

Which gives better value for money, GPT-5.6 Luna or SpaceXAI: Grok 4.6?

GPT-5.6 Luna. Cost per point divides the measured dollars LiveBench spent running the benchmark by the score it earned, so it captures something token pricing misses: a reasoning model can emit many times more tokens than its per-token rate suggests. GPT-5.6 Luna works out at $0.0911 per point and SpaceXAI: Grok 4.6 at $0.1181.

Does GPT-5.6 Luna or SpaceXAI: Grok 4.6 have a bigger context window?

GPT-5.6 Luna has the larger context window: 1.1M for GPT-5.6 Luna against 500K for SpaceXAI: Grok 4.6. Note that a window you can fill is not a window you should fill — retrieval quality usually degrades well before the limit, and you pay for every token you put in it.

Do GPT-5.6 Luna and SpaceXAI: Grok 4.6 support prompt caching?

Both publish a cached-input rate: $0.020 per million for GPT-5.6 Luna and $0.500 for SpaceXAI: Grok 4.6, against full input rates of $0.200 and $2.00. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.

Related comparisons

How these numbers are produced

  • Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
  • ScoresLiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
  • Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
  • “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.

Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.