ik_llama.cpp
Hacker ApprovedShares rank 9 of 27 inference engines by recommenders
Recommended as a faster, better-quantized fork of llama.cpp for running large models locally, especially on CPU and CUDA.
Recommended as a faster, better-quantized fork of llama.cpp for running large models locally, especially on CPU and CUDA.
13 people gave an opinion
- recommend it
- 12
- criticize it
- 1
- approval share
- 92%
Recommended in 9 of 13 months
- November 2025: recommended
- December 2025: recommended
- January 2026: not recommended
- February 2026: recommended
- March 2026: recommended
- April 2026: recommended
- May 2026: not recommended
- June 2026: recommended
- July 2026: recommended
- August 2026: recommended
- September 2026: recommended
- October 2026: not recommended
- November 2026: not recommended (month in progress)
Why it carries the badge
- 12 people recommend itNeeds 10 or more, met.
- 12 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 9 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
Better quants and quantization strategies
“IQK/Trellis quants from ik_llama.cpp and EXL3 quants from exllama should work better.”
Comment from 23 Aug 2026 in “Why your local LLM feels dumber than it is” “for Q4, unsloth XL or ik_llama IQ quants are usually better at the same size”
Comment from 1 Mar 2026 in “Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers” “ik_llama.cpp has more advanced CPU quantization strategies than llama.cpp”
Comment from 12 Aug 2026 in “Nvidia Nemotron 3.5 lightning and Nemo Switchyard” - Comment from 22 Mar 2026 in “Flash-Moe: Running a 397B Parameter Model on a Mac with 48GB RAM”
- Comment from 17 Apr 2026 in “Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to All”
Faster inference, especially when tuned
“A 5.5x perf increase was achieved by 1. recompiling with optimization flags 2. using ik_llama.cpp!”
Comment from 10 Mar 2026 in “AI on a Budget: Recompiling Llama.cpp for Qwen3.5 Inference on an HP Z440” “Inference speed is very good with ik_llama.cpp”
Comment from 8 Mar 2026 in “How to run Qwen 3.5 locally” “ik_llama is almost always faster when tuned”
Comment from 22 Dec 2025 in “A guide to local coding models” - Comment from 29 Apr 2026 in “HERMES.md: Anthropic bug causes $200 extra charge, refuses refund”
Makes large local models more practical
“ik_llama.cpp has some performance optimizations that make it semi-practical to run a model of this size on a server”
Comment from 6 Nov 2025 in “Kimi K2 Thinking, a SOTA open-source trillion-parameter reasoning model” “the ik fork has some enhancements that make this sort of thing more tolerable”
Comment from 1 Dec 2025 in “DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning”
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 | — | — |
| ik_llama.cppThis page | 12 | 1 | 92% | 9 | — | Approved |
What people said
18 of the 19 comments that mention it, newest first. Each links to the original comment.
“which have very good support for CUDA and CPU”
“Q8_0 from llama.cpp / ik_llama.cpp should also work better than FP8 from vllm”
“IQK/Trellis quants from ik_llama.cpp and EXL3 quants from exllama should work better.”
“ik_llama.cpp has more advanced CPU quantization strategies than llama.cpp”
“ik_llama.cpp loses quality for me”
“it amazes me how fast ik_llama.cpp moves”
“It works, thanks to https://github.com/ikawrakow/ik_llama.cpp/pull/1911 , which got merged in early June.”
“Hello, it's the internet calling, today is that day.”
“for better performance than llama”
“he has made multiple new and better quants”
“SOTA dynamic quants (especially ik_llama.cpp)”
“A 5.5x perf increase was achieved by 1. recompiling with optimization flags 2. using ik_llama.cpp!”
“Inference speed is very good with ik_llama.cpp”
“for Q4, unsloth XL or ik_llama IQ quants are usually better at the same size”
“ik_llama has prompt caching which help speed this up when swapping between agent contexts”
“ik_llama is almost always faster when tuned”
“the ik fork has some enhancements that make this sort of thing more tolerable”
“ik_llama.cpp has some performance optimizations that make it semi-practical to run a model of this size on a server”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.