Gemini 3 Flash: what 31 developers actually say.

Model · Updated Nov 2026

31 Hacker News opinions from 27 developers, Nov 2025 – Nov 2026. 26 come from people who say they use it. Here is what they praise, what they complain about, and who switched.

31 opinions27 developersNov 2025 – Nov 2026Methodology

Approval

85%

±10.5 at 90% confidence

Last 90 days

–

No recent data

Rank

of 45 models · CI-adjusted

Opinions

26 first-hand

Developers

1.1 opinions each

The short version

The case for Gemini 3 Flash, in one paragraph.

Gemini 3 Flash earns 85% approval across 31 opinions. Developers are divided on it.

See comparisons →

Pick it if you

Want a low-cost daily coding model
Need fast code maps and context selection
Already work in Copilot, Junie, Roo Code or OpenCode

Look elsewhere if you

Need strict adherence to custom instructions
Want clean syntax on the first pass
Need a model that weighs the full design before coding

Where it wins and loses.

Mentions across 31 opinions. Bars are counts, not votes; a tool can be both praised and criticized on the same topic.

TopicComplaints ← → PraiseMentions

PraiseComplaint

0 topics shown · sorted by net sentiment · source: 31 opinions

View data table

Topic	Praise	Complaints	Mentions

How opinion has moved.

Monthly approval, Dec 2025 – May 2026. Bars show opinion volume; low-volume months are still included but faded.

67% Dec 202567% Dec 2025 low100% May 2026– last 90 days

Monthly approval, 0–100. Bars show opinion volume. Months with fewer than 3 opinions are left out.

View data table

Month	Approval	Opinions
Dec 2025	67%	3
Jan 2026	71%	7
Feb 2026	94%	9
Mar 2026	88%	4
Apr 2026	90%	5
May 2026	100%	3

What developers say.

Summaries are paraphrased. Each links to the original Hacker News comment.

Praise

4 selected

Code quality

Considers Gemini 3 Flash the sweet spot in Junie, matching JetBrains' own suggestion.

HN comment · May 20, 2026Read on HN ↗

Autonomy

Gemini 3 Flash also worked for bug hunting on simple targets, used through Copilot.

HN comment · May 14, 2026Read on HN ↗

Cost & pricing

Gemini 3 Flash is a cost-effective model they liked after testing via Roo Code.

HN comment · May 3, 2026Read on HN ↗

They use Gemini 3 Flash as their main model in a custom agent harness that built their AI streamer platform.

HN comment · Apr 12, 2026Read on HN ↗

Complaints

2 selected

Code quality

Tried Gemini 3 Flash several times and found its output consistently useless.

HN comment · Jan 15, 2026Read on HN ↗

Customization

Gemini 3 Flash ignores their custom instructions on programming and Linux questions, unlike other models.

HN comment · Jan 12, 2026Read on HN ↗

Questions developers ask.

Is Gemini 3 Flash worth it?

Most commenters say yes. It holds 85% approval from 31 opinions, and cost/pricing drew 6 positive mentions with no negatives. A few users call its output useless or too costly, so results vary by setup.

Is Gemini 3 Flash better than GPT-5.2?

The data is thin. One commenter preferred GPT-5.2 instead, and two said they preferred Gemini 3 Flash over GitHub Copilot. There is not enough head-to-head evidence to call a winner.

Is Gemini 3 Flash good for coding?

Code quality is its strongest aspect by volume, with 8 positive mentions and 1 negative. Users praise it for bug hunting, Lua work, and main-model use in custom agents. Critics say it often writes buggy code and skips parts of prompts.

The data behind this page.

Developers give Gemini 3 Flash an 85% approval score across 31 opinions from 27 authors, and 26 of those opinions are firsthand. Approval rose from 67% in December 2025 to 94% in February 2026, then held between 88% and 100% through May 2026, though the later months rest on 3 to 5 opinions each. Code quality drew 8 positive comments against 1 negative, cost drew 6 positive and speed 3, with users calling it a cheap daily driver inside Copilot, Junie, Roo Code and OpenCode. Complaints focus on buggy output, broken syntax, ignored custom instructions, and a habit of jumping to a solution before reading the design problem.

Every number comes from public Hacker News comments collected through the HN Search API. A language model tagged each comment for sentiment, first-hand use, topic and switches; a human spot-checks samples. No vendor can pay for placement.

20,064 comments screened31 about Gemini 3 Flash26 first-hand

Read the full methodology →