cd /news/artificial-intelligence/solving-hack-the-box-challenges-with… · home topics artificial-intelligence article
[ARTICLE · art-97755] src=theaq.blog ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Solving Hack The Box Challenges with DeepSeek V4 Pro

DeepSeek V4 Pro solved 8 of 16 Hack The Box challenges in the HTB-Challenger Benchmark, scoring 36.2% with no false positives, but its median cost per challenge was $0.35, comparable to GPT-5.6 Luna despite being 17 times more expensive per test. The model, released four months ago, lagged behind benchmark leaders, highlighting rapid progress in AI capabilities.

read4 min views3 publishedAug 12, 2026
Solving Hack The Box Challenges with DeepSeek V4 Pro
Image: Theaq (auto-discovered)

DeepSeek V4 Pro was the best (and most expensive) LLM I tested with Strix a few months ago. After the heartbreaking results of its smaller sibling, DeepSeek V4 Flash 0731, in my HTB-Challenger tests, I was curious to see how the Pro version would handle the new challenges.

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.

The results of DeepSeek V4 Pro in my testing were, well, not great and not terrible. On the one hand, it didn’t have the basic flaws I observed when testing DeepSeek V4 Flash 0731: it didn’t submit any incorrect flags, and it used the available tools without any issues. On the other hand, its final score was comparable to that of GPT-5.6 Luna, whose median cost per test was 17 times lower.

The performance gap between DeepSeek V4 Pro and the leaders of my benchmark was also quite large. I’m afraid that’s a sign of just how quickly things have progressed over the last few months. DeepSeek V4 Pro was released four months ago, which feels like an eternity now that we’re getting better and better models almost every month.

Overall benchmark results #

Number of challenges: 16Number of solved challenges: 8Number of false positives: 0Runs where the model gave up: 0Runs that reached the step or cost limit: 8Runs where the model got stuck: 0Benchmark score: 36.2%

Metric Per challenge (median) Total
Model steps 62.5 938
Model cost $0.35 $11.84
Duration 00:06:19 03:34:35
Number of input tokens 1.93M 39.64M
Number of output tokens 0.01M 0.35M
Number of read_file tool calls
1.5 55
Number of write_file tool calls
1.0 92
Number of execute_command tool calls
26.5 821
Number of web_search tool calls
0.0 102

Results by challenge difficulty #

Very Easy challenges

Number of challenges: 4Number of solved challenges: 3Number of false positives: 0Runs where the model gave up: 0Runs that reached the step or cost limit: 1Runs where the model got stuck: 0Benchmark score: 71.4%

Metric Per challenge (median) Total
Model steps 24.5 158
Model cost $0.11 $1.19
Duration 00:03:51 00:27:35
Number of input tokens 0.28M 4.28M
Number of output tokens 0.01M 0.05M
Number of read_file tool calls
1.5 8
Number of write_file tool calls
0.0 7
Number of execute_command tool calls
15.5 149
Number of web_search tool calls
2.0 10

Easy challenges

Number of challenges: 4Number of solved challenges: 3Number of false positives: 0Runs where the model gave up: 0Runs that reached the step or cost limit: 1Runs where the model got stuck: 0Benchmark score: 72.0%

Metric Per challenge (median) Total
Model steps 21.5 155
Model cost $0.09 $2.09
Duration 00:04:05 00:49:04
Number of input tokens 0.22M 5.02M
Number of output tokens 0.01M 0.07M
Number of read_file tool calls
2.0 10
Number of write_file tool calls
2.0 41
Number of execute_command tool calls
15.5 79
Number of web_search tool calls
4.5 32

Medium challenges

Number of challenges: 4Number of solved challenges: 2Number of false positives: 0Runs where the model gave up: 0Runs that reached the step or cost limit: 2Runs where the model got stuck: 0Benchmark score: 49.0%

Metric Per challenge (median) Total
Model steps 56.5 225
Model cost $0.61 $3.78
Duration 00:09:55 01:11:26
Number of input tokens 2.44M 10.07M
Number of output tokens 0.02M 0.11M
Number of read_file tool calls
0.5 17
Number of write_file tool calls
0.5 5
Number of execute_command tool calls
52.5 231
Number of web_search tool calls
0.0 8

Hard challenges

Number of challenges: 4Number of solved challenges: 0Number of false positives: 0Runs where the model gave up: 0Runs that reached the step or cost limit: 4Runs where the model got stuck: 0Benchmark score: 0.0%

Metric Per challenge (median) Total
Model steps 100 400
Model cost $1.27 $4.78
Duration 00:16:12 01:06:30
Number of input tokens 5.25M 20.27M
Number of output tokens 0.02M 0.10M
Number of read_file tool calls
1.0 20
Number of write_file tool calls
4.0 39
Number of execute_command tool calls
93.5 362
Number of web_search tool calls
0.0 52
── more in #artificial-intelligence 4 stories · sorted by recency
── more on @deepseek v4 pro 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:4min 2026-08-12 ·