Ministral
Hacker ApprovedRanked 4 of 10 small language models by recommenders
Small free models people run locally for fast data extraction and simple tasks; critics say they trail larger models and are weak at vision.
14 people recommended Ministral in Hacker News comments, 3 had mixed feelings and 5 criticized it, November 2025 to October 2026.
Summary written by Claude from 28 comments. It was written before the November 2026 recount, and 26 of those comments still count.
22 people gave an opinion
- recommend it
- 14
- mixed feelings
- 3
- warn against or regret it
- 5
The record
Recommended in 7 of 12 months
In 14 different threads.
- November 2025: not recommended
- December 2025: recommended
- January 2026: recommended
- February 2026: recommended
- March 2026: recommended
- April 2026: not recommended
- May 2026: recommended
- June 2026: recommended
- July 2026: recommended
- August 2026: not recommended
- September 2026: not recommended
- October 2026: not recommended
Why it carries the badge
- 14 people recommend itNeeds 10 or more, met.
- 2.8 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 7 of 12 monthsNeeds 4 or more, met.
- Recommended in 14 threadsNeeds 5 or more, met.
- The biggest thread supplies 21% of its recommendersNeeds half or less, met.
The same rules apply to every product, and the badge can’t be bought. Mixed feelings count toward neither side. How the badge works
What people say about it
Reasons grouped from a sample of 26 of the 28 opinions about it, picked in proportion to how many people recommend it, have mixed feelings and criticize it. Each mark is one comment in the sample. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Fast responses for small tasks
“It's faster than other models of similar size I used previously, fast enough that I never wait for it”
Comment from 28 Feb 2026 in “How do I cancel my ChatGPT subscription?” “I was pleasantly surprised by how fast and accurate these small models have become”
Comment from 8 Feb 2026 in “Show HN: LocalGPT – A local-first AI assistant in Rust with persistent memory” “its actually much faster than Google/DDG searches and I use it a ton when coding”
Comment from 18 May 2026 in “Mistral's CEO: Europe has 2 years to stop becoming America's AI 'vassal state'” - Comment from 27 Apr 2026 in “France's Mistral Built a $14B AI Empire by Not Being American”
Runs locally on modest hardware
“still running Qwen and Ministral and their like on local hardware”
Comment from 31 Jan 2026 in “Code is cheap. Show me the talk” “I run Ministral-3b on my hardware as my go-to”
Comment from 22 May 2026 in “The current AI pricing was always going to go away” “Works fine on gaming PCs.”
Comment from 2 Jan 2026 in “Ask HN: Is the window for local-first AI closing?” Accurate and consistent results
“I have processed several hundred receipts this way with very high accuracy”
Comment from 21 Dec 2025 in “Structured outputs create false confidence” “it performed the best in my benchmarks”
Comment from 16 Mar 2026 in “Mistral Releases Leanstral” Useful for receipts and data extraction
“Works really well. Extracts companies you have dealt with, people, topics, events, locations, financial transactions, bills, etc.”
Comment from 17 Mar 2026 in “Mistral Releases Leanstral”
What people warn about
Each mark is one comment
Trails larger US and Chinese models
“Looks like another Deepseek distil like the new Ministrals. For every other use case that would be an insult”
Comment from 9 Dec 2025 in “Mistral releases Devstral2 and Mistral Vibe CLI” “Disappointed in Mistral/Ministral being so far behind these US & Chinese models”
Comment from 2 Apr 2026 in “Google releases Gemma 4 open models” “it's far worse at following instructions and providing entertaining content”
Comment from 3 Dec 2025 in “Mistral 3 family of models released” Misses details and makes errors
“didn't catch anything unless I gave some hints”
Comment from 11 Feb 2026 in “The Day the Telnet Died”
Other small language models people recommend
The four most recommended other small language models, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Mixed | Criticize | Approval | Months | Had for | Badge |
|---|---|---|---|---|---|---|---|
| OpenAI Luna | 171 | 31 | 46 | 79% | 4 | — | Approved |
| Claude Haiku | 147 | 48 | 113 | 57% | 12 | — | — |
| Apple Foundation Models | 24 | 17 | 31 | 44% | 10 | — | — |
| MinistralThis page | 14 | 3 | 5 | 74% | 7 | — | Approved |
| LFM2.5 2.6B | 11 | 1 | 8 | 58% | 6 | — | — |
What people said
26 of the 28 opinions about it, picked in proportion to the three counts, newest first. Each links to the original comment.
“boring models (Ministral 3, Mistral Small 4, Small 3, Small 3.1”
“a new Ministral model would also hit hard there”
“even a model like ministral-3:8b or qwen3.5:9b is a boon for so many smaller use cases”
“I run Ministral-3b on my hardware as my go-to”
“its actually much faster than Google/DDG searches and I use it a ton when coding”
“punches above its weight class”
“they're quite mediocre at vision so i haven't found much use for this”
“Disappointed in Mistral/Ministral being so far behind these US & Chinese models”
“Works really well. Extracts companies you have dealt with, people, topics, events, locations, financial transactions, bills, etc.”
“it performed the best in my benchmarks”
“Financial data extraction now works OK ish”
“It's faster than other models of similar size I used previously, fast enough that I never wait for it”
“I tested locally using Ministral 3B and it was promising.”
“didn't catch anything unless I gave some hints”
“I was pleasantly surprised by how fast and accurate these small models have become”
“found it very useful in low-risk environments”
“still running Qwen and Ministral and their like on local hardware”
“I'm optimistic about ministral3:3b”
“Works fine on gaming PCs.”
“I have processed several hundred receipts this way with very high accuracy”
“Looks like another Deepseek distil like the new Ministrals. For every other use case that would be an insult”
“it's far worse at following instructions and providing entertaining content”
“still very flakey w/ multi-turn”
“Honestly ministral would probably be sufficient if you can tolerate a small failure rate.”
“There's a very cool demo of that here”
“hopefully Ministral 14B lives up to their benchmarks”
Counts come from 1,791,911 Hacker News comments labeled by Claude, November 2025 to October 2026, with data through October 31, 2026. Counts updated 2 Nov 2026. How we count. Report a mistake.