---
type: "post"
id: "on-public-cyberbench-deepseek-v4-1-flash-scores-db4c38"
slug: "on-public-cyberbench-deepseek-v4-1-flash-scores-db4c38"
title: "On public CyberBench , DeepSeek V4.1 Flash scores 73.69% at $0.07/test vs Sol's 77.98% at $1.34/test."
description: "CyberBench v1.1 reports a measurable cost–accuracy tradeoff on the task of finding and patching security bugs."
retrieval_nugget: "CyberBench v1.1 reports a measurable cost–accuracy tradeoff on the task of finding and patching security bugs."
published_at: "2026-10-20"
updated_at: "2026-10-20"
record_date: "2026-10-20"
date_kind: "published_at"
topics: ["security"]
source_urls: ["https://www.vals.ai/benchmarks/cyber"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:on-public-cyberbench-deepseek-v4-1-flash-scores-db4c38"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:27f83b23d3a00f5c95a3029ef8eae2a4dbb096efdf0b13b3aa0fbe85f972a6c7","reviewed_at":"2026-10-20T11:04:21.795Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":1,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Benchmark utility for security agents depends on reading cost and task performance together, rather than treating a leaderboard rank as a deployment recommendation.","observed_facts":[{"text":"CyberBench v1.1 is described as a benchmark for finding and patching security bugs.","source_urls":["https://www.vals.ai/benchmarks/cyber"]},{"text":"DeepSeek V4.1 Flash scores 73.69% accuracy at $0.07 per test.","source_urls":["https://www.vals.ai/benchmarks/cyber"]},{"text":"GPT-6 Sol scores 77.98% accuracy at $1.34 per test.","source_urls":["https://www.vals.ai/benchmarks/cyber"]}],"mechanism":"Paired accuracy and per-test cost figures show the price of the incremental benchmark performance, enabling teams to assess whether it matters for their workload.","why_now":"CyberBench v1.1 presents these measures side by side for a current set of models, making the tradeoff legible to teams comparing security-task candidates.","implications":["Teams should validate both bug-finding accuracy and per-task cost on representative repositories before choosing a model."],"evidence_boundary":"The supplied source describes benchmark scores and costs only; it does not establish production performance, patch correctness in a team’s environment, or generalization beyond CyberBench.","watch_conditions":["Whether subsequent evaluations report reproducible results across repository types and separate finding quality from patch quality."],"related_records":[],"new_branch_reason":"No prior New Runtime analytical record was supplied; this begins a distinct record on the benchmark cost–accuracy tradeoff for security-task models."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/on-public-cyberbench-deepseek-v4-1-flash-scores-db4c38/","markdown":"https://newruntime.com/posts/on-public-cyberbench-deepseek-v4-1-flash-scores-db4c38.md","json":"https://newruntime.com/posts/on-public-cyberbench-deepseek-v4-1-flash-scores-db4c38.json"}
related_patterns: ["ai-engineering-evaluations-are-splitting-into-task-specific","agent-economics-moves-to-completed-work"]
---

# On public CyberBench , DeepSeek V4.1 Flash scores 73.69% at $0.07/test vs Sol's 77.98% at $1.34/test.

## Retrieval answer

CyberBench v1.1 reports a measurable cost–accuracy tradeoff on the task of finding and patching security bugs.

CyberBench v1.1 reports a measurable cost–accuracy tradeoff on the task of finding and patching security bugs. DeepSeek V4.1 Flash scores 73.69% at $0.07 per test; GPT-6 Sol scores 77.98% at $1.34 per test. The listed costs make GPT-6 Sol roughly 19 times more expensive per test, for a 4.29-point accuracy advantage.

Thesis: Benchmark utility for security agents depends on reading cost and task performance together, rather than treating a leaderboard rank as a deployment recommendation. Mechanism: Paired accuracy and per-test cost figures show the price of the incremental benchmark performance, enabling teams to assess whether it matters for their workload. Why now: CyberBench v1.1 presents these measures side by side for a current set of models, making the tradeoff legible to teams comparing security-task candidates.

First implication: Teams should validate both bug-finding accuracy and per-task cost on representative repositories before choosing a model. Evidence boundary: The supplied source describes benchmark scores and costs only; it does not establish production performance, patch correctness in a team’s environment, or generalization beyond CyberBench. Watch condition: Whether subsequent evaluations report reproducible results across repository types and separate finding quality from patch quality.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Security](https://newruntime.com/topics/security.md): Explore the security topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)