HWBenchedge-AI hardware data, cited & graded

What is the fastest local LLM hardware under $2,000?

Intel 265K Custom PC (Nvidia RTX 4090) (as tested by Jeff Geerling): 335 tok/s (Llama 3.2:3b, as in linked issue (Ollama default tags are 4-bit); grade C; undated). Next: Intel 13900K (Nvidia 5090) (as tested by Jeff Geerling) 271 tok/s; Intel Core Ultra 7 368H 250 tok/s. Like-for-like group: LLM ≤4B, batch 1.

Top results
#PlatformResultPerf/WPriceGrade · dateSource
1Intel 265K Custom PC (Nvidia RTX 4090) (as tested by Jeff Geerling)335 tok/s
Llama 3.2:3b · as in linked issue (Ollama default tags are 4-bit)
1.28 measuredunknownC undatedJeff Geerling ai-benchmarks README (Findings tables)
2Intel 13900K (Nvidia 5090) (as tested by Jeff Geerling)271 tok/s
Llama 3.2:3b · as in linked issue (Ollama default tags are 4-bit)
unknownunknownD undatedJeff Geerling ai-benchmarks README (Findings tables)
3Intel Core Ultra 7 368H250 tok/s
minicpm4-0.5b · INT4-MIXED
unknownunknownD undatedOpenVINO official LLM benchmarks (llm_models_7-368H.csv)
4Ryzen 9 7900X (Nvidia 4090) (as tested by Jeff Geerling)237 tok/s
Llama 3.2:3b · as in linked issue (Ollama default tags are 4-bit)
unknownunknownD undatedJeff Geerling ai-benchmarks README (Findings tables)
5Intel 13900K (Nvidia 4090) (as tested by Jeff Geerling)216 tok/s
Llama 3.2:3b · as in linked issue (Ollama default tags are 4-bit)
unknownunknownD undatedJeff Geerling ai-benchmarks README (Findings tables)

Filters: task=LLM generation, budget=2000. Ranking: boards that meet every known target, then headroom and perf/W; unknown power or price is not penalised. Open in /find · JSON

Sources

Generated 2026-10-16 from HWBench data only. All answers · Submit a result

Machine-readable: /find.json (see llms.txt); full tables are in “Show data”. Confidence grades: A audited MLPerf closed · B documented third-party · C community · D vendor-published · E estimate. Unknown values are shown as unknown, never estimated. Generated 2026-10-16 from public sources.