{"slug": "solving-hack-the-box-challenges-with-deepseek-v4-pro", "title": "Solving Hack The Box Challenges with DeepSeek V4 Pro", "summary": "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.", "body_md": "# Solving Hack The Box Challenges with DeepSeek V4 Pro\n\nDeepSeek V4 Pro was the best (and most expensive) LLM I tested with [Strix a few months ago](/2026/04/25/deepseek-v4-with-strix-a-quick-test.html). After the heartbreaking results of its smaller sibling, [DeepSeek V4 Flash 0731](/2026/08/11/solving-htb-challenges-with-deepseek-deepseek-v4-flash-0731.html), in my HTB-Challenger tests, I was curious to see how the Pro version would handle the new challenges.\n\nThis blog post is part of a series of tests for the HTB-Challenger Benchmark. See the\n\n[benchmark results page]for all results and the[benchmark methodology]to learn how the benchmark is calculated.\n\nThe 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](/2026/08/10/solving-htb-challenges-with-openai-gpt-5.6-luna.html), whose median cost per test was 17 times lower.\n\nThe 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.\n\n## Overall benchmark results\n\n**Number of challenges:** 16**Number of solved challenges:** 8**Number of false positives:** 0**Runs where the model gave up:** 0**Runs that reached the step or cost limit:** 8**Runs where the model got stuck:** 0**Benchmark score:** 36.2%\n\n| Metric | Per challenge (median) | Total |\n|---|---|---|\n| Model steps | 62.5 | 938 |\n| Model cost | $0.35 | $11.84 |\n| Duration | 00:06:19 | 03:34:35 |\n| Number of input tokens | 1.93M | 39.64M |\n| Number of output tokens | 0.01M | 0.35M |\nNumber of `read_file` tool calls |\n1.5 | 55 |\nNumber of `write_file` tool calls |\n1.0 | 92 |\nNumber of `execute_command` tool calls |\n26.5 | 821 |\nNumber of `web_search` tool calls |\n0.0 | 102 |\n\n## Results by challenge difficulty\n\n### Very Easy challenges\n\n**Number of challenges:** 4**Number of solved challenges:** 3**Number of false positives:** 0**Runs where the model gave up:** 0**Runs that reached the step or cost limit:** 1**Runs where the model got stuck:** 0**Benchmark score:** 71.4%\n\n| Metric | Per challenge (median) | Total |\n|---|---|---|\n| Model steps | 24.5 | 158 |\n| Model cost | $0.11 | $1.19 |\n| Duration | 00:03:51 | 00:27:35 |\n| Number of input tokens | 0.28M | 4.28M |\n| Number of output tokens | 0.01M | 0.05M |\nNumber of `read_file` tool calls |\n1.5 | 8 |\nNumber of `write_file` tool calls |\n0.0 | 7 |\nNumber of `execute_command` tool calls |\n15.5 | 149 |\nNumber of `web_search` tool calls |\n2.0 | 10 |\n\n### Easy challenges\n\n**Number of challenges:** 4**Number of solved challenges:** 3**Number of false positives:** 0**Runs where the model gave up:** 0**Runs that reached the step or cost limit:** 1**Runs where the model got stuck:** 0**Benchmark score:** 72.0%\n\n| Metric | Per challenge (median) | Total |\n|---|---|---|\n| Model steps | 21.5 | 155 |\n| Model cost | $0.09 | $2.09 |\n| Duration | 00:04:05 | 00:49:04 |\n| Number of input tokens | 0.22M | 5.02M |\n| Number of output tokens | 0.01M | 0.07M |\nNumber of `read_file` tool calls |\n2.0 | 10 |\nNumber of `write_file` tool calls |\n2.0 | 41 |\nNumber of `execute_command` tool calls |\n15.5 | 79 |\nNumber of `web_search` tool calls |\n4.5 | 32 |\n\n### Medium challenges\n\n**Number of challenges:** 4**Number of solved challenges:** 2**Number of false positives:** 0**Runs where the model gave up:** 0**Runs that reached the step or cost limit:** 2**Runs where the model got stuck:** 0**Benchmark score:** 49.0%\n\n| Metric | Per challenge (median) | Total |\n|---|---|---|\n| Model steps | 56.5 | 225 |\n| Model cost | $0.61 | $3.78 |\n| Duration | 00:09:55 | 01:11:26 |\n| Number of input tokens | 2.44M | 10.07M |\n| Number of output tokens | 0.02M | 0.11M |\nNumber of `read_file` tool calls |\n0.5 | 17 |\nNumber of `write_file` tool calls |\n0.5 | 5 |\nNumber of `execute_command` tool calls |\n52.5 | 231 |\nNumber of `web_search` tool calls |\n0.0 | 8 |\n\n### Hard challenges\n\n**Number of challenges:** 4**Number of solved challenges:** 0**Number of false positives:** 0**Runs where the model gave up:** 0**Runs that reached the step or cost limit:** 4**Runs where the model got stuck:** 0**Benchmark score:** 0.0%\n\n| Metric | Per challenge (median) | Total |\n|---|---|---|\n| Model steps | 100 | 400 |\n| Model cost | $1.27 | $4.78 |\n| Duration | 00:16:12 | 01:06:30 |\n| Number of input tokens | 5.25M | 20.27M |\n| Number of output tokens | 0.02M | 0.10M |\nNumber of `read_file` tool calls |\n1.0 | 20 |\nNumber of `write_file` tool calls |\n4.0 | 39 |\nNumber of `execute_command` tool calls |\n93.5 | 362 |\nNumber of `web_search` tool calls |\n0.0 | 52 |", "url": "https://wpnews.pro/news/solving-hack-the-box-challenges-with-deepseek-v4-pro", "canonical_source": "https://theaq.blog/2026/08/12/solving-htb-challenges-with-deepseek-deepseek-v4-pro.html", "published_at": "2026-08-12 09:11:15+00:00", "updated_at": "2026-08-15 07:43:13.532928+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research"], "entities": ["DeepSeek V4 Pro", "DeepSeek V4 Flash 0731", "GPT-5.6 Luna", "Hack The Box", "HTB-Challenger Benchmark"], "alternates": {"html": "https://wpnews.pro/news/solving-hack-the-box-challenges-with-deepseek-v4-pro", "markdown": "https://wpnews.pro/news/solving-hack-the-box-challenges-with-deepseek-v4-pro.md", "text": "https://wpnews.pro/news/solving-hack-the-box-challenges-with-deepseek-v4-pro.txt", "jsonld": "https://wpnews.pro/news/solving-hack-the-box-challenges-with-deepseek-v4-pro.jsonld"}}