Hardware /

RTX 4050 Laptop GPU

234measured results
155best gen tok/s*
6 GBreported memory

*Headline excludes tiny/test models (under 0.6B params or TinyStories-style). Open in rig checker →

llama.cpp

ModelQuantGen tok/sPrompt tok/sSourceContext testedBuild
Sailor-0.5B-ChatQ4_K - Medium1559,049LocalScore ↗512+128llamafile 0.9.3Details →
Qwen2.5 Coder 0.5B InstructQ6_K1429,277LocalScore ↗512+128llamafile 0.9.3Details →
AMD OLMo 1B SFT DPOQ4_K - Medium1396,524LocalScore ↗512+128llamafile 0.9.3Details →
LamoFast_2.0Q8_01349,362LocalScore ↗512+128llamafile 0.9.3Details →
Llama 3.2 1B InstructQ4_K - Medium1325,332LocalScore ↗512+128llamafile 0.9.3Details →
Llama 3.2 1B InstructQ4_K - Medium1325,231LocalScore ↗4,352llamafile 0.9.3Details →
Qwen3-0.6BQ4_K - Medium1226,559LocalScore ↗512+128llamafile 0.9.3Details →
LightOnOCR 2 1BQ4_K - Medium1196,786LocalScore ↗512+128llamafile 0.9.3Details →
Beck 0.6BQ4_K - Medium1186,774LocalScore ↗512+128llamafile 0.9.3Details →
ReasonLite 0.6B TurboQ4_K - Medium1176,679LocalScore ↗512+128llamafile 0.9.3Details →
InternVL3_5 1BQ4_K - Medium1166,705LocalScore ↗512+128llamafile 0.9.3Details →
Theta Crucis 0.6B Turbo1Q5_K - Medium1156,533LocalScore ↗512+128llamafile 0.9.3Details →
Qwen3 0.6BQ4_K - Medium1156,639LocalScore ↗512+128llamafile 0.9.3Details →
Llama 3.2 1B InstructQ4_K - Medium1144,979LocalScore ↗512+128llamafile 0.9.2Details →
ReasonLite 0.6BQ4_K - Medium1146,885LocalScore ↗512+128llamafile 0.9.3Details →
Astral 0.6B CoderQ5_K - Medium1136,593LocalScore ↗512+128llamafile 0.9.3Details →
Deepseek Coder 1.3b InstructQ4_K - Medium1075,212LocalScore ↗512+128llamafile 0.9.3Details →
MinerU HTMLQ4_K - Medium1066,591LocalScore ↗512+128llamafile 0.9.3Details →
DeepBrainz R1 0.6B v2Q4_K - Medium1066,629LocalScore ↗512+128llamafile 0.9.3Details →
Titulm Mpt 1b v2.0Q4_K - Medium1024,778LocalScore ↗512+128llamafile 0.9.3Details →
Gemma 3 1b It QatQ4_01008,463LocalScore ↗512+128llamafile 0.9.3Details →
tinyllama_tinyllama-1.1b-chat-v1.0Q8_099.53,987LocalScore ↗512+128llamafile 0.9.3Details →
Gemma-3-1B-ItQ4_K - Medium98.18,031LocalScore ↗512+128llamafile 0.9.3Details →
Nusantara 0.8b Indo ChatQ4_K - Medium94.25,652LocalScore ↗512+128llamafile 0.9.3Details →
Llama 3.2 1B InstructQ4_K - Medium94.14,127LocalScore ↗4,352llamafile 0.9.3Details →
Refact 1_6B FimQ4_K - Medium92.93,189LocalScore ↗512+128llamafile 0.9.3Details →
ordis_v355_VarGH_mergedQ4_K - Medium92.84,816LocalScore ↗512+128llamafile 0.9.3Details →
Qwen3 1.7BQ4_K - Medium91.03,512LocalScore ↗4,352llamafile 0.9.3Details →
DeepSeek R1 Distill Qwen 1.5BQ4_K - Medium90.34,795LocalScore ↗512+128llamafile 0.9.3Details →
Gemma 3 1b ItQ4_K - Medium87.08,007LocalScore ↗512+128llamafile 0.9.3Details →
VibeThinker 1.5BQ5_K - Medium86.14,694LocalScore ↗512+128llamafile 0.9.3Details →
OpenCoder 1.5B InstructQ4_K - Medium82.14,476LocalScore ↗512+128llamafile 0.9.3Details →
OpenCoder 1.5B InstructQ4_K - Medium81.24,249LocalScore ↗512+128llamafile 0.9.3Details →
Qwen3 Coder 0.6B 0.9B 7 1bR5 Nobs F16Q4_K - Medium79.04,600LocalScore ↗512+128llamafile 0.9.3Details →
Sailor2 1B ChatQ4_K - Medium78.54,527LocalScore ↗512+128llamafile 0.9.3Details →
Superthoughts Lite v2 MOE Llama3.2Q4_K - Medium78.22,911LocalScore ↗512+128llamafile 0.9.3Details →
Kakaocorp.Kanana 1.5 2.1b Instruct 2505Q4_K - Medium73.13,037LocalScore ↗512+128llamafile 0.9.3Details →
EXAONE 3.5 2.4B InstructQ4_K - Medium72.22,756LocalScore ↗512+128llamafile 0.9.3Details →
Neo 3 3B A400M BaseQ4_K - Medium68.42,324LocalScore ↗512+128llamafile 0.9.3Details →
Fijik1.5 2.6b A380MQ4_K - Medium66.22,472LocalScore ↗512+128llamafile 0.9.3Details →
TowerVision 2BQ4_K - Medium65.43,533LocalScore ↗512+128llamafile 0.9.3Details →
Granite 3.1 1b A400M InstructQ5_K - Medium62.13,457LocalScore ↗512+128llamafile 0.9.3Details →
Llama 3.2 3B InstructQ4_K - Medium60.02,518LocalScore ↗512+128llamafile 0.9.3Details →
Starcoder2 3bQ4_K - Medium59.02,377LocalScore ↗512+128llamafile 0.9.3Details →
Nanbeige4.1 3BQ4_K - Medium56.72,332LocalScore ↗512+128llamafile 0.9.3Details →
Nanbeige4 3B Thinking 2511Q4_K - Medium55.82,387LocalScore ↗512+128llamafile 0.9.3Details →
Phi 4 Mini InstructQ4_K - Medium50.52,065LocalScore ↗512+128llamafile 0.9.3Details →
Phi-4-Mini-ReasoningQ4_K - Medium50.12,066LocalScore ↗512+128llamafile 0.9.3Details →
EuroMoE 2.6B A0.6B Instruct PreviewQ4_K - Medium49.01,635LocalScore ↗512+128llamafile 0.9.3Details →
Gemma-3-4B-ItQ4_K - Medium46.72,551LocalScore ↗512+128llamafile 0.9.3Details →
Rio 3.0 Open MiniQ4_K - Medium45.81,785LocalScore ↗512+128llamafile 0.9.3Details →
Qwen3-4B-Instruct-2507Q4_K - Medium45.61,804LocalScore ↗512+128llamafile 0.9.3Details →
Gemma 3 4b ItQ4_K - Medium45.52,455LocalScore ↗512+128llamafile 0.9.3Details →
Qwen3 4B Thinking 2507Q4_K - Medium45.31,806LocalScore ↗512+128llamafile 0.9.3Details →
Llama 2 7BQ4_043.71,726llama.cpp (CUDA) ↗512+128llama.cpp CUDA @d79d8f3Details →
Syngen Reasoning 0.6bBF1642.53,838LocalScore ↗512+128llamafile 0.9.3Details →
Llama 2 7BQ4_041.91,154llama.cpp (Vulkan) ↗512+128llama.cpp Vulkan @d79d8f3Details →
Meta Llama 3.1 8B InstructQ4_K - Medium31.71,217LocalScore ↗4,352llamafile 0.9.3Details →
Meta Llama 3.1 8B InstructQ4_K - Medium31.01,185LocalScore ↗512+128llamafile 0.9.3Details →
Nemotron Cascade 8BQ4_K - Medium30.11,132LocalScore ↗512+128llamafile 0.9.3Details →
KomdigiUB 8B Instruct DTPQ4_K - Medium29.91,151LocalScore ↗512+128llamafile 0.9.3Details →
DeepSeek R1 0528 Qwen3 8BQ4_K - Medium29.71,159LocalScore ↗512+128llamafile 0.9.3Details →
Falcon3 Mamba 7B InstructQ4_014.782.6LocalScore ↗512+128llamafile 0.9.3Details →
171 tiny/test model result(s) (excluded from headlines)
ModelQuantGen tok/sPrompt tok/sSourceContext testedBuild
Tiny Llama Ultra CompactQ4_K - Medium3,091220,409LocalScore ↗512+128llamafile 0.9.3Details →
Planck-OpenLAiNN-10MF162,811218,864LocalScore ↗512+128llamafile 0.9.3Details →
Planck-OpenLAiNN-10MQ4_02,330206,117LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Gpt2Q4_K - Medium2,032235,917LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ4_K - Medium1,905145,343LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ4_K - Medium1,890144,854LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ5_11,890145,629LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ5_K - Small1,890142,236LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ4_11,887148,849LocalScore ↗512+128llamafile 0.9.3Details →
Raincandy U TinyStories 656KQ4_K - Medium1,885153,940LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ4_K - Small1,885140,884LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ5_K - Medium1,879153,054LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ6_K1,873159,348LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ5_K - Medium1,869145,396LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ4_01,859147,634LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ5_01,856154,382LocalScore ↗512+128llamafile 0.9.3Details →
Raincandy U TinyStories 656KQ8_01,823153,794LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories-656KQ4_K - Medium1,732153,201LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Random Phi3ForCausalLMQ4_K - Medium1,686175,578LocalScore ↗512+128llamafile 0.9.3Details →
llama-39mQ4_K - Medium1,529124,862LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Random Olmo HfQ4_K - Medium1,492175,963LocalScore ↗512+128llamafile 0.9.3Details →
Tiny_Random_Llama2_NewQ4_K - Medium1,492167,051LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Mpt Random Remote CodeQ3_K - Medium1,42052,177LocalScore ↗512+128llamafile 0.9.3Details →
FalconQ4_K - Medium1,414101,933LocalScore ↗512+128llamafile 0.9.3Details →
vicuna-68mQ4_K - Medium1,37590,490LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Random OrionQ4_K - Medium1,169158,772LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Random Qwen3Q4_K - Medium958167,212LocalScore ↗512+128llamafile 0.9.3Details →
llamaall F32 (guessed)90892,104LocalScore ↗512+128llamafile 0.9.3Details →
Bloomall F32899135,614LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ3_K - Medium89838,782LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Random Qwen3Q4_K - Medium897157,670LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ3_K - Large86539,062LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ3_K - Medium85432,354LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Gemma TestQ4_K - Medium849129,655LocalScore ↗512+128llamafile 0.9.3Details →
llamaQ6_K84679,711LocalScore ↗512+128llamafile 0.9.3Details →
llamaQ4_K - Medium84487,067LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Random Internlm2Q4_K - Medium833124,180LocalScore ↗512+128llamafile 0.9.3Details →
Experimental Neo TinyStories Korean 800K 20240819Q4_K - Medium82474,343LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Lm 8MQ4_K - Medium80774,142LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KIQ3_S mix - 3.66 bpw72821,230LocalScore ↗512+128llamafile 0.9.3Details →
Tinyllama 15MQ4_K - Medium71874,446LocalScore ↗512+128llamafile 0.9.3Details →
NexTrim 50MQ4_K - Medium70459,232LocalScore ↗512+128llamafile 0.9.3Details →
Pythia 14mQ4_K - Medium697101,949LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories-656KIQ4_XS - 4.25 bpw67415,970LocalScore ↗512+128llamafile 0.9.3Details →
Tinyllama 15MQ4_K - Medium67377,040LocalScore ↗512+128llamafile 0.9.3Details →
TinyLLamaF1665738,404LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KIQ4_NL - 4.5 bpw65125,663LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ2_K - Medium64621,761LocalScore ↗512+128llamafile 0.9.3Details →
Pythia 14mQ4_K - Medium643101,046LocalScore ↗512+128llamafile 0.9.3Details →
Raincandy U TinyStories 656KQ2_K - Medium64027,724LocalScore ↗512+128llamafile 0.9.3Details →
Granite 3.1 2b InstructQ6_K63996,076LocalScore ↗512+128llamafile 0.9.3Details →
GPTNeoX Spanish_Poet 70mQ4_K - Medium62853,522LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KIQ4_XS - 4.25 bpw62530,058LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KIQ3_S - 3.4375 bpw61723,654LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KIQ3_XS - 3.3 bpw61724,298LocalScore ↗512+128llamafile 0.9.3Details →
Pythia 70m DedupedQ4_K - Medium61559,284LocalScore ↗512+128llamafile 0.9.3Details →
Distilgpt2 Tiny ConversationalQ6_K60540,369LocalScore ↗512+128llamafile 0.9.3Details →
TinyStories 656KQ3_K - Small55714,647LocalScore ↗512+128llamafile 0.9.3Details →
TinyLLama v0Q4_K - Medium52834,967LocalScore ↗512+128llamafile 0.9.3Details →
DAC60MQ4_K - Medium52545,270LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q4_052234,113LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q5_052132,770LocalScore ↗512+128llamafile 0.9.3Details →
TinyLLama v0Q4_K - Medium51933,597LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q5_151730,545LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q4_K - Small51732,351LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q4_K - Medium51332,676LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q4_151033,013LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q8_050933,036LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 04Q4_K - Medium50531,669LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 02Q4_K - Medium50430,373LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q6_K50232,163LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q5_K - Medium50231,486LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q4_K - Medium50233,450LocalScore ↗512+128llamafile 0.9.3Details →
Plyx 15MQ4_K - Medium48654,473LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q4_K - Medium47832,105LocalScore ↗512+128llamafile 0.9.3Details →
mistral-60mQ4_K - Medium47649,439LocalScore ↗512+128llamafile 0.9.3Details →
Config.JsonQ8_046995,601LocalScore ↗512+128llamafile 0.9.3Details →
Lite-Oute-1-65M-InstructQ4_K - Medium46929,126LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q4_045933,267LocalScore ↗512+128llamafile 0.9.3Details →
Pico-OpenLAiNN-100MF1644836,744LocalScore ↗512+128llamafile 0.9.3Details →
Phi 3 Mini 4k InstructQ4_K - Medium41457,990LocalScore ↗512+128llamafile 0.9.3Details →
Tiny MistralQ4_K - Medium40917,695LocalScore ↗512+128llamafile 0.9.3Details →
Tiny Random Stablelm 2Q4_K - Medium40729,877LocalScore ↗512+128llamafile 0.9.3Details →
Granite 3.0 1b A400M InstructQ6_K40064,106LocalScore ↗512+128llamafile 0.9.3Details →
AMD Llama 135mQ4_K - Medium37955,309LocalScore ↗512+128llamafile 0.9.3Details →
AMD Llama 135mQ4_K - Medium37852,402LocalScore ↗512+128llamafile 0.9.3Details →
Stentor 12M InstructQ4_K - Medium36355,775LocalScore ↗512+128llamafile 0.9.3Details →
Mamba Mistral Instruct DistillQ4_K - Medium34921,636LocalScore ↗512+128llamafile 0.9.3Details →
Out_Bak_100M_2K_Code_Iter_160000Q4_K - Medium33723,097LocalScore ↗512+128llamafile 0.9.3Details →
Amd Pretrained 135m 2k Slimpajama_No_Book3Q4_K - Medium32322,291LocalScore ↗512+128llamafile 0.9.3Details →
Pythia 160m DedupedQ4_K - Medium32322,273LocalScore ↗512+128llamafile 0.9.3Details →
Gpt2Q4_K - Medium30121,300LocalScore ↗512+128llamafile 0.9.3Details →
LocutusqueXFelladrin TinyMistral248M InstructQ6_K29910,880LocalScore ↗512+128llamafile 0.9.3Details →
Jeeves Small 100MQ4_K - Medium29023,345LocalScore ↗512+128llamafile 0.9.3Details →
TinyMistral 248M v2.5Q6_K28611,023LocalScore ↗512+128llamafile 0.9.3Details →
RainQ4_028222,390LocalScore ↗512+128llamafile 0.9.3Details →
MiniPLM Qwen 200MQ6_K28021,889LocalScore ↗512+128llamafile 0.9.3Details →
llamaIQ4_XS - 4.25 bpw27913,888LocalScore ↗512+128llamafile 0.9.3Details →
Llama 2 Tiny 4kv Heads 16layers RandomQ4_K - Medium26010,922LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03BF1623923,089LocalScore ↗512+128llamafile 0.9.3Details →
Qwen2 96MQ4_K - Medium23419,873LocalScore ↗512+128llamafile 0.9.3Details →
Jais Family 256m ChatQ4_K - Medium23011,486LocalScore ↗512+128llamafile 0.9.3Details →
OpenELM 270M InstructQ4_K - Medium22012,968LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ4_K - Medium21910,670LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ2_K - Medium2119,347LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ2_K - Small2108,760LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ4_120610,482LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ5_020410,493LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ4_020410,661LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ3_K - Small2049,703LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ4_K - Small20310,347LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ5_K - Small20210,384LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ3_K - Medium20110,136LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ5_K - Medium20010,035LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ4_K - Medium19810,530LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ3_K - Large19610,154LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ5_119610,424LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ6_K1899,569LocalScore ↗512+128llamafile 0.9.3Details →
OpenELM 270MQ4_K - Medium18712,828LocalScore ↗512+128llamafile 0.9.3Details →
OuteTTS 0.1 350M HfQ4_K - Medium18710,662LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatunknown, may not work17510,676LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q3_K - Medium1751,976LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatQ8_01739,982LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q3_K - Large1722,096LocalScore ↗512+128llamafile 0.9.3Details →
Llm 350m Instruct v2Q4_K - Medium1699,257LocalScore ↗4,352llamafile 0.9.3Details →
Gemma 3 270m Qat Q4_0 UnquantizedQ4_016823,746LocalScore ↗512+128llamafile 0.9.3Details →
Functiongemma-270M-ItQ4_K - Medium16623,516LocalScore ↗512+128llamafile 0.9.3Details →
DogeAI v1.5 CoderQ4_K - Medium16323,266LocalScore ↗512+128llamafile 0.9.3Details →
Gemma 3 270m It Qat Q4_0 UnquantizedQ4_016324,023LocalScore ↗512+128llamafile 0.9.3Details →
Gemma-3-270M-ItQ4_K - Medium15923,440LocalScore ↗512+128llamafile 0.9.3Details →
MobileLLM 80M FinetunedQ6_K15314,090LocalScore ↗512+128llamafile 0.9.3Details →
MobileLLM 125M HFQ4_K - Medium15313,033LocalScore ↗512+128llamafile 0.9.3Details →
SmolLM2 135M InstructQ4_015213,164LocalScore ↗512+128llamafile 0.9.3Details →
SmollerLM-48M-Instruct-ft-sftQ4_K - Medium14614,471LocalScore ↗512+128llamafile 0.9.3Details →
SmolVLM2 256M Video InstructQ4_K - Medium14612,890LocalScore ↗512+128llamafile 0.9.3Details →
SmolLM2 135M InstructQ4_K - Medium14513,269LocalScore ↗512+128llamafile 0.9.3Details →
DMindAI.DMind 3 NanoQ4_K - Medium14321,525LocalScore ↗512+128llamafile 0.9.3Details →
Distil Home Assistant FunctiongemmaF1613922,376LocalScore ↗512+128llamafile 0.9.3Details →
Gemma 3 270m ItQ4_K - Medium13823,291LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatF161369,907LocalScore ↗512+128llamafile 0.9.3Details →
SmolDocling 256M PreviewQ4_K - Medium13513,054LocalScore ↗512+128llamafile 0.9.3Details →
SmolLM2 135M InstructQ4_K - Medium12913,112LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03IQ4_NL - 4.5 bpw1281,546LocalScore ↗512+128llamafile 0.9.3Details →
OpenCLAW SEED 135MQ4_K - Medium12713,254LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03IQ4_XS - 4.25 bpw1251,483LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q3_K - Small1251,431LocalScore ↗512+128llamafile 0.9.3Details →
Minueza 2 96M Instruct Variant 03Q2_K - Medium1231,464LocalScore ↗512+128llamafile 0.9.3Details →
MTLM1 200MQ4_K - Medium1228,946LocalScore ↗512+128llamafile 0.9.3Details →
SmolVLM2 500M Video InstructQ4_K - Medium1217,382LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatall F3280.36,559LocalScore ↗512+128llamafile 0.9.3Details →
mamba-130m-hfQ4_K - Medium79.9881LocalScore ↗512+128llamafile 0.9.3Details →
MonadQ4_K - Medium75.912,204LocalScore ↗512+128llamafile 0.9.3Details →
MiniPLM Mamba 130MQ4_K - Medium75.81,053LocalScore ↗512+128llamafile 0.9.3Details →
Mamba 130m HfQ4_K - Small74.2824LocalScore ↗512+128llamafile 0.9.3Details →
Gemma-3-270M-It-QatIQ4_XS - 4.25 bpw59.31,111LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatBF1647.85,258LocalScore ↗512+128llamafile 0.9.3Details →
DistilGPT OSS Qwen3 4BQ4_K - Medium45.21,763LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ1_S - 1.5625 bpw38.6150LocalScore ↗512+128llamafile 0.9.3Details →
State Spaces Mamba 370m HfQ4_K - Medium38.1400LocalScore ↗512+128llamafile 0.9.3Details →
SmolLM2 135M InstructIQ1_S - 1.5625 bpw35.4739LocalScore ↗512+128llamafile 0.9.3Details →
SmolLM2 135M InstructIQ4_XS - 4.25 bpw35.0860LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ2_XS - 2.3125 bpw34.9143LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ2_M - 2.7 bpw30.5135LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ1_M - 1.75 bpw30.494.5LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ3_S mix - 3.66 bpw30.3122LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ2_S - 2.5 bpw30.1131LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ2_XXS - 2.0625 bpw28.9138LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ4_XS - 4.25 bpw28.4329LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ3_XXS - 3.0625 bpw26.3120LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ4_NL - 4.5 bpw26.3259LocalScore ↗512+128llamafile 0.9.3Details →
h2o-danube3-500m-chatIQ3_S - 3.4375 bpw24.994.5LocalScore ↗512+128llamafile 0.9.3Details →

Ollama

No measured Ollama results yet. Submit yours →

MLX

No measured MLX results yet. Submit yours →

Includes results from LocalScore, a Mozilla Builders project by cjpais, mirrored in the cn0303/local-llm-speed-benchmarks dataset. LocalScore runs use llamafile, which is built on llama.cpp.