DevConsensus
Models / DeepSeek V4 Flash
Model · HN opinions Nov 2025 – Nov 2026

DeepSeek V4 Flash: what developers actually say

We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on DeepSeek V4 Flash. 129 of the 151 opinions come from people who say they use it themselves.

83%approval, all opinions
80%approval, last 90 days
#1of 45 models ranked
120different developers

Developers rate DeepSeek V4 Flash highly: 83% approval across 151 opinions, and 129 of those opinions come from direct use. Cost is the strongest theme (55 positive mentions against 4 negative), ahead of speed (29 to 4) and code quality (28 to 2). Approval over the last 90 days is 80%, below the overall figure, and September 2026 dropped to 72% on 25 opinions after 86% in August. Complaints cluster around latency through OpenRouter and reliability, which drew 9 positive mentions against 5 negative.

Pick it if

  • Want low cost per task
  • Want fast output and solid code quality
  • Want to keep work on your own hardware

Look elsewhere if

  • Need steady latency through a third-party router
  • Depend on strict tool schemas and style control
  • Want the newest DeepSeek release

Strengths and weaknesses

Each row counts opinions that mentioned the topic. Bars to the right of the line are praise, bars to the left are complaints.

PraiseComplaints
Cost & pricing
4
55
Speed
4
29
Code quality
2
28
Reliability
5
9
Privacy & local use
11
Autonomy
1
8

How opinion has moved

050100Apr '26May '26Jun '26Jul '26Aug '26Sep '26
Monthly approval (line) and number of opinions (bars). Months with fewer than 3 opinions are left out.

What developers say

Praise

Considers DeepSeek V4 Flash much smarter than MiMo 2.5.HN comment · Sep 23, 2026
Says DeepSeek V4 Flash can follow a handwritten spec and find flaws in the implementation.HN comment · Sep 23, 2026
Runs DeepSeek V4 Flash with 1M context on a GB10 Spark and says it codes very well.HN comment · Sep 21, 2026
Says DeepSeek V4 Flash is their daily driver and remarkably capable for its cost.HN comment · Sep 21, 2026
Says DeepSeek V4 Flash was cheap enough to get their side-project work done.HN comment · Sep 19, 2026
Says DeepSeek V4 Flash was already quite good before the V4.1 release.HN comment · Sep 16, 2026
Says DeepSeek V4 Flash is more proficient and efficient than Claude at code review, and cheaper to run.HN comment · Sep 10, 2026
Codes side projects almost exclusively with DeepSeek V4 Flash, which recovers well from errors.HN comment · Sep 9, 2026

Complaints

Found DeepSeek costlier per message than Luna, likely due to long thinking times.HN comment · Sep 29, 2026
Says DeepSeek V4 Flash ignored a custom tool's schema and used a different argument name.HN comment · Sep 17, 2026
Says DeepSeek V4 Flash was too slow to be usable on their local hardware and lost to Qwen.HN comment · Sep 10, 2026
Says DeepSeek V4 Flash latency on OpenRouter is unusable.HN comment · Sep 9, 2026
Says DeepSeek V4 Flash is not much better than GLM at following style instructions; calls it stubborn.HN comment · Sep 8, 2026
Says DeepSeek V4 Flash via OpenRouter was unusable, with provider switching, cache resets, and rate limits.HN comment · Aug 28, 2026
Reports DeepSeek V4 Flash 0731 produced walls of gibberish in Hermes when reasoning was enabled.HN comment · Aug 23, 2026
Says DeepSeek V4 Flash tells them to check pages manually rather than verifying UI changes with screenshots.HN comment · Aug 21, 2026

Summaries are paraphrased. Each links to the original comment.

FAQ

Is DeepSeek V4 Flash worth it?

Most developers say yes. Approval sits at 83% from 151 opinions, and cost is the most cited reason, with 55 positive mentions against 4 negative. Some users still find it pricier per message than cheaper alternatives because of long thinking times.

Is DeepSeek V4 Flash good for coding?

Code quality drew 28 positive mentions and only 2 negative. Users report it follows a written spec, spots flaws in an implementation, and recovers from errors. A few note hallucinations and early stops, so review still matters.

Is DeepSeek V4 Flash better than GLM?

The data is split. Two people preferred it over GLM 5.2, while two named GLM 5.3 Flash as what they use instead. Three others moved to DeepSeek V4.1 Flash, so version choice matters as much as brand.