Gemma 4 26B inference engine for Mac
Shares rank 6 of 27 inference engines by recommenders
Free open-source engine that runs Gemma 4 26B in about 2 GB of RAM on M-series Macs. Critics say generation is too slow on some Macs.
Free open-source engine that runs Gemma 4 26B in about 2 GB of RAM on M-series Macs. Critics say generation is too slow on some Macs.
17 people gave an opinion
- recommend it
- 15
- criticize it
- 2
- approval share
- 88%
Recommended in 1 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: 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.
- 7.5 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 1 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
Usable token speeds on Apple Silicon
“Got 48 tok/s decode at 1.9 GB RSS (2.4 GB peak), faster than the 24 GB M5 Pro mentioned in the benchmarks.”
Comment from 30 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” “12 tok/s and almost instant response in M1 Max Mac Studio (with faster SSD than laptops) are impressive”
Comment from 29 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” “I tried it on my wife's M1 MacBook Air 512GB and it gets 4–5 tok/s”
Comment from 29 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” - Comment from 30 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac”
Large model runs in 2 GB RAM
“Got 48 tok/s decode at 1.9 GB RSS (2.4 GB peak), faster than the 24 GB M5 Pro mentioned in the benchmarks.”
Comment from 30 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” “seeing 14GB of weights crammed into 2GB of RAM is impressive.”
Comment from 29 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” Impressive, does what it says
“It does exactly what it says it does.”
Comment from 29 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” “Tried it and i'm impressed.”
Comment from 31 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac”
What people warn about
Each mark is one comment
Too slow on some Macs
“the token generation rate is simply far too low”
Comment from 30 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” “I'm only getting < 1 token/second on my Neo”
Comment from 30 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac”
Other inference engines people recommend
The four most recommended other inference engines, 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 |
|---|---|---|---|---|---|---|
| vLLM | 112 | 56 | 67% | 12 | — | Approved |
| Strata | 31 | 15 | 67% | 5 | — | Approved |
| DwarfStar 4 | 26 | 2 | 93% | 6 | — | Approved |
| SGLang | 19 | 13 | 59% | 8 | — | — |
| Gemma 4 26B inference engine for MacThis page | 15 | 2 | 88% | 1 | — | — |
What people said
15 of the 21 comments that mention it, newest first. Each links to the original comment.
“Tried it and i'm impressed.”
“I'm only getting < 1 token/second on my Neo”
“Got 48 tok/s decode at 1.9 GB RSS (2.4 GB peak), faster than the 24 GB M5 Pro mentioned in the benchmarks.”
“the token generation rate is simply far too low”
“Confirmed on my Neo! Got 4.5 tokens / second sustained.”
“Hell yea. ... It's a beautiful thing.”
“enjoy offline inference (and a bit of Gemma4 intelligence) for your everyday tasks”
“12 tok/s and almost instant response in M1 Max Mac Studio (with faster SSD than laptops) are impressive”
“You’re a mad man - thank you!”
“The two project might potentially work well together.”
“It does exactly what it says it does.”
“seeing 14GB of weights crammed into 2GB of RAM is impressive.”
“I tried it on my wife's M1 MacBook Air 512GB and it gets 4–5 tok/s”
“my engine may be a good fit”
“Nice job implementing expert caching!”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.