Local AI · Creator

antirez

19 published recipes.

Recipes

2× DGX Spark

DeepSeek-V4-Flash 0731 Q2 with DwarfStar on 2× DGX Spark (network TP)

by antirez

26tok/sEveryday

1 request, prose prompt (105 tokens), 64K allocated, no speculative decoding Source

Intelligence34with thinkingOriginal model's value. Heavily compressed (~2.1 bit), intelligence is likely much lower.

≤ 3 bit: quantization may cost quality

IQ2_XXS gate/up + Q2_K down (routed, imatrix), Q8 Attn/Shared/Out; Experten hälftig je Spark (~45 GiB je Host)DwarfStar (ds4) main@fc80bd69 (2026-10-07)Weights Repo Updated Nov 2, 2026

Details

1× DGX Spark

DeepSeek-V4-Flash IQ2_XXS with DwarfStar on 1× DGX Spark

by antirez

19.5tok/sEveryday

1 request, prose prompt, context 4096, no speculative decoding Source

Intelligence34with thinkingOriginal model's value. Heavily compressed (~2.1 bit), intelligence is likely much lower.

≤ 3 bit: quantization may cost quality

IQ2_XXS gate/up + Q2_K down (routed, imatrix), Q8 Attn/Shared/OutDwarfStar (ds4) main@0aaea5a (2026-09-20)Weights Repo Updated Nov 2, 2026

Details

1× DGX Spark

Qwen3.8-Flash-Next IQ2_XXS with DwarfStar on 1× DGX Spark

by antirez

17.1tok/sEveryday

1 request, prose prompt, context 8192, no speculative decoding Source

Intelligence40with thinkingOriginal model's value. Heavily compressed (~2.1 bit), intelligence is likely much lower.

≤ 3 bit: quantization may cost quality
License unclear

IQ2_XXS gate/up, Q2_K down; 41.73 GiB resident, 95.37 GiB BF16-n-grams auf SSDDwarfStar (ds4) main@0aaea5a (2026-09-20)Weights Repo Updated Nov 2, 2026

Details

1× Mac

GLM-5.3 UD-IQ2_XXS with DwarfStar on 1× Mac

by antirez

6.14tok/sMixed

1 request, mixed prompt set, context 1k, no speculative decoding Source

Intelligence45with thinkingOriginal model's value. Heavily compressed (~2.1 bit), intelligence is likely much lower.

≤ 3 bit: quantization may cost quality
Streams from SSD
License unclear

UD-IQ2_XXS routed, blk78 Q2K, 196.58 GiBDwarfStar (ds4) main@0aaea5a (2026-09-20)Weights Repo Updated Nov 2, 2026

Details

2× DGX Spark

DeepSeek-V4-Flash 0731 MXFP4 with DwarfStar + DSpark on 2× DGX Spark (network TP)

by antirez

75.6tok/sPeak

1 request, code prompt, DSpark speculative decoding, 64K allocated Source

Intelligence34with thinking

MXFP4 Experten (nativ), F16 HC/Compressor/Indexer, Q8 Attn/Shared/Out; Experten hälftig je Spark (~77 GiB je Host)DwarfStar (ds4) main@fc80bd69 (2026-10-07)Weights Repo Updated Nov 2, 2026

Details

1× Mac

Qwen3.8-Flash-Next 176.9B IQ2_XXS with DwarfStar on 1× Mac

by antirez

48.7tok/sPeak

1 request, best value reported by source, context 32k Source

Intelligence40with thinkingOriginal model's value. Heavily compressed (~2.1 bit), intelligence is likely much lower.

≤ 3 bit: quantization may cost quality
License unclear

IQ2_XXS gate/up, Q2_K down; 41.73 GiB resident, 95.37 GiB BF16-n-grams auf SSDDwarfStar (ds4) main@0aaea5a (2026-09-20)Weights Repo Updated Nov 2, 2026

Details

1× Mac

Qwen3.8-Flash-Next 176.9B Q4_K with DwarfStar on 1× Mac

by antirez

48.4tok/sPeak

1 request, best value reported by source, context 32k Source

Intelligence40with thinking

License unclear

Q4_K; 69.74 GiB resident, 95.37 GiB BF16-n-grams auf SSDDwarfStar (ds4) main@0aaea5a (2026-09-20)Weights Repo Updated Nov 2, 2026

Details

1× Mac

DeepSeek-V4-Flash 284B IQ2_XXS with DwarfStar on 1× Mac

by antirez

39.4tok/sPeak

1 request, best value reported by source, context 2k Source

Intelligence34with thinkingOriginal model's value. Heavily compressed (~2.1 bit), intelligence is likely much lower.

≤ 3 bit: quantization may cost quality

IQ2_XXS gate/up + Q2_K down (routed, imatrix), Q8 Attn/Shared/Out, F16/F32 RestDwarfStar (ds4) main@0aaea5a (2026-09-20)Weights Repo Updated Nov 2, 2026

Details

1× Mac