Leanstral by Mistral
The most recommended of 3 math reasoning models
Open-weight Lean proof model people can run locally on consumer hardware. Main caveat: critics find it trails larger models on coding and proof tasks.
Open-weight Lean proof model people can run locally on consumer hardware. Main caveat: critics find it trails larger models on coding and proof tasks.
27 people gave an opinion
- recommend it
- 15
- criticize it
- 12
- approval share
- 56%
Recommended in 2 of 13 months
- November 2025: not recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: recommended
- April 2026: not recommended
- May 2026: not recommended
- June 2026: not recommended
- July 2026: recommended
- August 2026: not recommended
- September 2026: not recommended
- October 2026: not recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 15 people recommend itNeeds 10 or more, met.
- 1.2 recommenders for every criticNeeds 2 to 1 or better, not met.
- Recommended in 2 of 13 monthsNeeds 4 or more, not met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Open weights, runs locally
“open weights + relatively small is still headline worthy. This thing really cooks.”
Comment from 4 Jul 2026 in “Leanstral 1.5: Proof Abundance for All” “We can run Leanstral locally on commodity/consumer hardware.”
Comment from 4 Jul 2026 in “Leanstral 1.5: Proof Abundance for All” “It looks like the weights are Apache-2.0 licensed.”
Comment from 16 Mar 2026 in “Mistral Releases Leanstral” Useful addition to proof toolkits
“would be a powerful complement to data‑driven model discovery”
Comment from 17 Mar 2026 in “Mistral Releases Leanstral” “Try out Leanstral 1.5 on the latest version of OpenATP!”
Comment from 4 Jul 2026 in “Leanstral 1.5: Proof Abundance for All” “this sounds like a great tool to add to the toolbelt”
Comment from 4 Jul 2026 in “Leanstral 1.5: Proof Abundance for All”
What people warn about
Each mark is one comment
Trails larger models on coding
“This model is specifically trained on this task and significantly[1] underperforms opus.”
Comment from 16 Mar 2026 in “Mistral Releases Leanstral” “I found it worse than gemma4 at coding even though it's 4x the parameter size”
Comment from 1 Jul 2026 in “Leanstral 1.5” “Haiku kinda sucks at this task, and Leanstral is worse?”
Comment from 16 Mar 2026 in “Mistral Releases Leanstral” Benchmark examples seem cherry-picked
“I don't think I'd consider that such a smashing success that it's worth bringing up as the sole example tbh”
Comment from 4 Jul 2026 in “Leanstral 1.5: Proof Abundance for All” “This is nice work, but I found the bug finding example to be weird”
Comment from 3 Jul 2026 in “Leanstral 1.5: Proof Abundance for All” “something felt cherry-picked”
Comment from 1 Jul 2026 in “Leanstral 1.5” Weak docs, FAQ and marketing
“Non-competitive products and models combined with bad marketing and GTM”
Comment from 1 Jul 2026 in “Leanstral 1.5” “just a half-assed FAQ, that seems vibe-coded and searched poorly”
Comment from 1 Jul 2026 in “Leanstral 1.5” “I'm amazed at how they manage to fumble releases over and over”
Comment from 1 Jul 2026 in “Leanstral 1.5” - Comment from 17 Mar 2026 in “Mistral Releases Leanstral”
Access restricted behind GitHub or Labs
“enabling Labs models isn't available for self-serve activation on standard individual accounts”
Comment from 1 Jul 2026 in “Leanstral 1.5” “I must connect to GitHub to use”
Comment from 1 Jul 2026 in “Leanstral 1.5”
Other math reasoning models people recommend
The two most recommended other math reasoning 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 | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| Leanstral by MistralThis page | 15 | 12 | 56% | 2 | — | — |
| DeepSeekMath-V2 | 10 | 6 | 63% | 6 | — | — |
| Harmonic Aristotle | 10 | 2 | 83% | 4 | — | Approved |
What people said
24 of the 27 comments that mention it, newest first. Each links to the original comment.
“We can run Leanstral locally on commodity/consumer hardware.”
“open weights + relatively small is still headline worthy. This thing really cooks.”
“this sounds like a great tool to add to the toolbelt”
“I don't think I'd consider that such a smashing success that it's worth bringing up as the sole example tbh”
“Try out Leanstral 1.5 on the latest version of OpenATP!”
“This is nice work, but I found the bug finding example to be weird”
“something felt cherry-picked”
“what they can accomplish with that lower scale, and it's stuff like Leanstral, Voxtral and other niche products”
“enabling Labs models isn't available for self-serve activation on standard individual accounts”
“I must connect to GitHub to use”
“Non-competitive products and models combined with bad marketing and GTM”
“These guys don't answer emails.”
“I'm amazed at how they manage to fumble releases over and over”
“just a half-assed FAQ, that seems vibe-coded and searched poorly”
“I found it worse than gemma4 at coding even though it's 4x the parameter size”
“basically not explaining at all what the setup does”
“would be a powerful complement to data‑driven model discovery”
“I don’t know that I’d want to have a system vibe code a proof.”
“Leanstral seems interesting, so we'll have to see how it does”
“it clearly and demonstrably does not”
“It looks like the weights are Apache-2.0 licensed.”
“This is great, congratulations to the Mistral team!”
“This model is specifically trained on this task and significantly[1] underperforms opus.”
“Haiku kinda sucks at this task, and Leanstral is worse?”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.