Liquid AI 8B-A1B MoE
Ranked 7 of 11 small language models by recommenders
People recommend it for fast local inference, including on CPUs and integrated GPUs. The most common caveats are that it is not very smart, weak at world knowledge, and below expectations.
People recommend it for fast local inference, including on CPUs and integrated GPUs. The most common caveats are that it is not very smart, weak at world knowledge, and below expectations.
19 people gave an opinion
- recommend it
- 12
- criticize it
- 7
- approval share
- 63%
Recommended in 5 of 13 months
- November 2025: not recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: not recommended
- April 2026: not recommended
- May 2026: recommended
- June 2026: recommended
- July 2026: recommended
- August 2026: recommended
- September 2026: not recommended
- October 2026: recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 12 people recommend itNeeds 10 or more, met.
- 1.7 recommenders for every criticNeeds 2 to 1 or better, not met.
- Recommended in 5 of 13 monthsNeeds 4 or more, 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
Very fast local inference
“Bad at translation, at least to Russian. Very fast though, about 2x faster than Gemma 4 e2b on my CPU.”
Comment from 30 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T” “it's my go-to model for whenever i need fast local inference”
Comment from 1 Jul 2026 in “Liquid AI releases a 230M model optimized for phones, Raspberry Pi, and robots” “I liked LFM2-8B-A1B for its speed on cpu or integrated gpu”
Comment from 2 May 2026 in “LFM2-24B-A2B: Scaling Up the LFM2 Architecture” - Comment from 29 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T”
- Comment from 16 Jun 2026 in “Running local models is good now”
- Comment from 29 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T”
- Comment from 29 May 2026 in “LFM2.5-8B-A1B: An Even Better On-Device Mixture-of-Experts”
- Comment from 29 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T”
Runs on CPU or integrated GPU
“Bad at translation, at least to Russian. Very fast though, about 2x faster than Gemma 4 e2b on my CPU.”
Comment from 30 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T” “I liked LFM2-8B-A1B for its speed on cpu or integrated gpu”
Comment from 2 May 2026 in “LFM2-24B-A2B: Scaling Up the LFM2 Architecture” “Guess we can run this even on CPU!”
Comment from 29 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T” Useful for focused, single-purpose tasks
“I’ve also found LFM2.5 8B surprisingly useful for single-focus tasks”
Comment from 24 Aug 2026 in “Anthropic's best AI model struggles to attract users as cheaper tools thrive” “the Liquid AI models are already very good at a lot of this”
Comment from 29 Oct 2026 in “Introducing LFM3: Every Modality for Specialized Intelligence” “it executed it extremely well”
Comment from 29 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T”
What people warn about
Each mark is one comment
Not very smart, limited knowledge
“it's expectedly not favorable for world knowledge”
Comment from 30 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T” “these things clearly don't know much yet”
Comment from 29 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T” “It's not very smart, to say the least.”
Comment from 24 Aug 2026 in “Anthropic's best AI model struggles to attract users as cheaper tools thrive” Below expectations on performance
“I just assumed it would perform better, given all the advancements in the space.”
Comment from 30 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T” “I was never able to get anything useful out of them”
Comment from 25 Mar 2026 in “Local LLM by Ente” “It did not perform as well as I expected.”
Comment from 29 May 2026 in “Liquid AI reveals 8B-A1B MoE trained on 38T”
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 | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| Claude Haiku | 218 | 134 | 62% | 13 | 4 mo4 owners | — |
| OpenAI Luna | 218 | 56 | 80% | 7 | — | Approved |
| Apple Foundation Models | 45 | 35 | 56% | 11 | 1 yr4 owners | — |
| Ministral | 18 | 6 | 75% | 8 | — | Approved |
| Liquid AI 8B-A1B MoEThis page | 12 | 7 | 63% | 5 | — | — |
What people said
21 of the 22 comments that mention it, newest first. Each links to the original comment.
“the Liquid AI models are already very good at a lot of this”
“It's not very smart, to say the least.”
“I’ve also found LFM2.5 8B surprisingly useful for single-focus tasks”
“it's my go-to model for whenever i need fast local inference”
“seem to both be very high token rate”
“I just assumed it would perform better, given all the advancements in the space.”
“it has significant restrictions for any org with other $10M in revenue”
“it's expectedly not favorable for world knowledge”
“Bad at translation, at least to Russian. Very fast though, about 2x faster than Gemma 4 e2b on my CPU.”
“it is borderline incapable of using tool calls”
“these things clearly don't know much yet”
“It did not perform as well as I expected.”
“I really love how fast it is!”
“I kinda feel like they are overtraining their models”
“its a very small model, and very fast”
“Guess we can run this even on CPU!”
“it executed it extremely well”
“I'm particularly excited for this one as it may allow teams to scale this architecture for VLAs”
“pretty speedy for local compute!”
“I liked LFM2-8B-A1B for its speed on cpu or integrated gpu”
“I was never able to get anything useful out of them”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.