cd /news/artificial-intelligence/solving-hack-the-box-challenges-with… · home topics artificial-intelligence article
[ARTICLE · art-94385] src=theaq.blog ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Solving Hack The Box Challenges with GPT-5.6 Luna Pro

OpenAI's GPT-5.6 Luna Pro solved 11 of 16 Hack The Box challenges, scoring 55.0% on the HTB-Challenger Benchmark, compared with GPT-5.6 Luna which solved only one Medium and no Hard challenges. The model, served via OpenRouter.ai with reasoning.mode set to 'pro', is approximately five times more expensive than Luna but remains cheap, with a median cost of $0.10 per challenge and total cost of $3.05. It solved all Very Easy and Easy challenges, 2 of 4 Medium, and 1 of 4 Hard, with no false positives.

read5 min views1 publishedAug 10, 2026
Solving Hack The Box Challenges with GPT-5.6 Luna Pro
Image: source

What the heck is GPT-5.6 Luna Pro, I hear you asking. That’s a very good question! To answer it, let me quote its description on OpenRouter.ai: “* GPT-5.6 Luna Pro is the same underlying model as GPT-5.6 Luna, served with *”

reasoning.mode

set to pro

for higher-quality responses on complex tasks.Okay, so how much better, and how much more expensive, is it compared with GPT-5.6 Luna, I hear you asking now. And that’s exactly what I can tell you. This blog post is part of a series of tests for the HTB-Challenger Benchmark. See the

[benchmark results page]for all results and the[benchmark methodology]to learn how the benchmark is calculated.

I tested GPT-5.6 Luna in my previous post, so I have the numbers ready. In a nutshell: Luna Pro is approximately five times more expensive thanLuna. It may sound dramatic, but** Lunais such a cheap model that Luna Prois still very cheap.- In this benchmark run, Luna Pro performed much better thanLuna**. If you look below at the results for the more difficult challenges, you can see that** Luna Prosolved some Medium and Hard challenges, while Luna**solved only one Medium challenge and no Hard challenges.

I believe the main use case for Luna Pro is if you are tied to the OpenAI ecosystem and need a model for offensive-security or CTF-style tasks. In my tests, GPT-5.6 Terra and GPT-5.6 Sol refused to respond to most offensive-security requests, so GPT-5.6 Luna Pro was the most useful OpenAI option for this kind of work.

Overall benchmark results #

Number of challenges: 16Number of solved challenges: 11Number of false positives: 0Runs where the model gave up: 4Runs that reached the step or cost limit: 1Runs where the model got stuck: 0Benchmark score: 55.0%

Metric Per challenge (median) Total
Model steps 21.5 554
Model cost $0.10 $3.05
Duration 00:08:55 03:48:35
Number of input tokens 1.50M 42.93M
Number of output tokens 0.07M 1.63M
Number of read_file tool calls
4.0 186
Number of write_file tool calls
0.0 18
Number of execute_command tool calls
36.5 865
Number of web_search tool calls
0.0 3

Results by challenge difficulty #

Very Easy challenges

Number of challenges: 4Number of solved challenges: 4Number of false positives: 0Runs where the model gave up: 0Runs that reached the step or cost limit: 0Runs where the model got stuck: 0Benchmark score: 100.0%

Metric Per challenge (median) Total
Model steps 6 38
Model cost $0.01 $0.14
Duration 00:01:43 00:12:53
Number of input tokens 0.16M 1.89M
Number of output tokens 0.01M 0.08M
Number of read_file tool calls
0.5 7
Number of write_file tool calls
0.0 1
Number of execute_command tool calls
6.0 73
Number of web_search tool calls
0.0 0

Easy challenges

Number of challenges: 4Number of solved challenges: 4Number of false positives: 0Runs where the model gave up: 0Runs that reached the step or cost limit: 0Runs where the model got stuck: 0Benchmark score: 100.0%

Metric Per challenge (median) Total
Model steps 21.5 80
Model cost $0.09 $0.34
Duration 00:08:46 00:37:25
Number of input tokens 1.38M 4.94M
Number of output tokens 0.06M 0.22M
Number of read_file tool calls
1.5 12
Number of write_file tool calls
0.0 0
Number of execute_command tool calls
26.5 99
Number of web_search tool calls
0.5 3

Medium challenges

Number of challenges: 4Number of solved challenges: 2Number of false positives: 0Runs where the model gave up: 2Runs that reached the step or cost limit: 0Runs where the model got stuck: 0Benchmark score: 50.0%

Metric Per challenge (median) Total
Model steps 33 175
Model cost $0.16 $1.02
Duration 00:13:37 01:10:21
Number of input tokens 2.30M 14.23M
Number of output tokens 0.10M 0.54M
Number of read_file tool calls
8.5 75
Number of write_file tool calls
0.5 3
Number of execute_command tool calls
47.5 265
Number of web_search tool calls
0.0 0

Hard challenges

Number of challenges: 4Number of solved challenges: 1Number of false positives: 0Runs where the model gave up: 2Runs that reached the step or cost limit: 1Runs where the model got stuck: 0Benchmark score: 25.0%

Metric Per challenge (median) Total
Model steps 73.5 261
Model cost $0.45 $1.55
Duration 00:28:35 01:47:55
Number of input tokens 6.36M 21.87M
Number of output tokens 0.22M 0.79M
Number of read_file tool calls
16.5 92
Number of write_file tool calls
2.5 14
Number of execute_command tool calls
121.0 428
Number of web_search tool calls
0.0 0
── more in #artificial-intelligence 4 stories · sorted by recency
── more on @openai 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/solving-hack-the-box…] indexed:0 read:5min 2026-08-10 ·