Cite

/llms.txt · /llms-full.txt · 111 pairs · 62 models · 45 chips

Measured pairs

1. 

   Qwen3.8-27B · 370 tok/s

   RTX 5090 · TensorFold 0.6.6 · 4-bit MLX format + DFlash2

   @Oluwaphilemon1

2. 

   Qwen3.5 2B · 351 tok/s

   RTX 5090 32GB · unknown · Q4_K_M GGUF

   @fillagrew

3. 

   Qwen3.8-27B · 272.8 tok/s

   RTX PRO 6000 Blackwell Max-Q · TensorFold · NVFP4

   @llm_frontier_jp

4. 

   GLM 5.3 Flash · 264 tok/s

   4× RTX PRO 6000 · TensorFold 0.6.2 · EXL3

   @KurcideGet

5. 

   Qwen3.8-27B · 258 tok/s

   RTX 5090 · ExLlamaV3 · EXL3 4-bit + DFlash2

   @Oluwaphilemon1

6. 

   GLM 5.3 Flash · 238 tok/s

   2× RTX PRO 6000 · TensorFold 0.6.5 · EXL3

   @KurcideGet

7. 

   Swift 1.5 Qwen3.8 Flash-Next · 220 tok/s

   RTX 4090 24GB · Strata · IQ2_XS GGUF

   @Knuckles_XBTGet

8. 

   Qwen3.8-Flash-Next · 195.8 tok/s

   RTX 5090 · Strata · IQ3_S

   @therealskrlrGet

9. 

   Qwen3.8-Flash-Next · 191 tok/s

   RTX 5090 · Strata · Q2_0 GSQ-RCO

   @orenoa7Get

10. 

    Qwen3.8-27B · 186 tok/s

    RTX 5090 · llama.cpp · GGUF IQ4_XS + MTP

    @Oluwaphilemon1Get

11. 

    Qwen3.8-27B · 177.6 tok/s

    RTX A5000 · TensorFold · NVFP4

    @Ja6ekGet

12. 

    Gemma-4 26B-A4B-it · 173 tok/s

    RTX 5090 32GB · unknown · Q4_K_M GGUF

    @fillagrew

13. 

    Qwen3.8-Flash-Next · 158.3 tok/s

    2× RTX 3090 24GB · Strata · IQ3_S

    @geldekiGet

14. 

    Qwen3.8-Flash-Next · 153.4 tok/s

    2× RTX 3090 24GB · Strata · IQ3_S

    @geldekiGet

15. 

    Qwen3.8-27B · 151.4 tok/s

    RTX PRO 6000 Blackwell Max-Q · vLLM · NVFP4

    @llm_frontier_jp

16. 

    Qwen3.8-Flash-Next · 142 tok/s

    RTX 5090 32GB · Strata · IQ3_S GGUF

    @arianpgGet

17. 

    Qwen3.8-27B · 139.3 tok/s

    RTX A5000 · vLLM · NVFP4

    @Ja6ek

18. 

    Qwen3.8-Flash-Next Coder · 127 tok/s

    RTX 5060 Ti 16GB + RTX 3070 8GB · Strata · IQ1_M

    @VectorCrossProdGet

19. 

    GLM 5.3 Flash · 127 tok/s

    4× DGX Spark · TensorFold · TensorFold

    @Tech2Wild

20. 

    Qwen3.8-Flash-Next · 126.5 tok/s

    M5 Max · MTPLX V2.11.3 · unknown

    @Youssofal_

21. 

    Qwen3.8-Flash-Next · 122.5 tok/s

    2× RTX 3090 24GB · Strata 0.1.39 · GSQ-RCO IQ3_S

    @geldeki

22. 

    GLM 5.3 Flash · 120 tok/s

    RTX PRO 6000 + 2× GB10 · TensorFold · EXL3 4bpw

    @hokazuya

23. 

    Qwen3.8-Flash-Next · 114.3 tok/s

    RTX 5090 · llama.cpp · UD-Q4_K_XL

    @mizuirotGet

24. 

    Qwen3.6-35B-A3B · 108.6 tok/s

    M4 Pro 64GB · Rapid-MLX · 4-bit

    @draslan_eth

25. 

    GLM 5.3 Flash · 104 tok/s

    4× DGX Spark · unknown · EXL3 4 bpw

    @0xSero

26. 

    Qwen3.8-Flash-Next · 104 tok/s

    RTX 5090 32GB · Strata · IQ3_S GGUF

    @DaSun64125381Get

27. 

    DeepSeek v4.1 Flash · 103 tok/s

    2× DGX Spark · vLLM · custom 2.70 bpw

    @sudoingXGet

28. 

    Qwen3.8-27B · 102.9 tok/s

    DGX Spark · TensorFold · MLX 4-bit g64 + DFlash2

    @Oluwaphilemon1HF

29. 

    Qwen3.8-Flash-Next · 100.6 tok/s

    RTX 3090 24GB · Strata · IQ2_XS GGUF

    @draslan_ethGet

30. 

    Qwen3.8-Flash-Next · 100 tok/s

    RTX 3090 24GB · Strata · IQ3_S

    @RubenTorneyGet

31. 

    Qwen3.8-Flash-Next · 99.9 tok/s

    2× RTX 5060 Ti 16GB · Strata · IQ3_S GSQ RCO GGUF

    @needmorevramGet

32. 

    Qwen3.6-35B-A3B · 94.4 tok/s

    2× DGX Spark · unknown · NVFP4

    @bonellisystems

33. 

    DeepSeek v4.1 Flash · 93.6 tok/s

    2× DGX Spark · TensorFold · TensorFold

    @Tech2WildGet

34. 

    Qwen3.8-Flash-Next · 92.8 tok/s

    RTX PRO 6000 · llama.cpp b11412 · Q4_XL

    @piyo621

35. 

    Qwen3.8-Flash-Next · 92 tok/s

    RTX 4090 · Strata · UD-IQ4_XS

    @dextuneGet

36. 

    Qwen3.8-27B · 89.9 tok/s

    M5 Pro 64GB · TensorFold · MLX 4-bit + DFlash2

    @aartiles24HF

37. 

    Gemma-4 26B-A4B-it · 88 tok/s

    Radeon RX 9060 XT 16GB · llama.cpp v0.6.0 · UD-Q4_K_XL

    @RadxaNaokiGet

38. 

    Qwen3.8-Flash-Next · 87 tok/s

    RTX 3090 24GB · Strata · IQ2_XS

    @superaleshaGet

39. 

    Qwen3.6-35B-A3B · 85.5 tok/s

    M4 Pro · rapid-mlx · 4-bit MLX

    @rapidmlxHF

40. 

    GLM 5.3 Flash · 85 tok/s

    2x DGX Spark · TensorFold · EXL3

    @MehdiHacks

41. 

    Qwen3.5-4B · 82.8 tok/s

    M4 Pro · rapid-mlx · 4-bit MLX

    @rapidmlx

42. 

    GLM 5.3 · 81.5 tok/s

    2× DGX Station GB300 · SGLang · NVFP4

    @SJRonanMD

43. 

    Qwen3.8-Flash-Next · 80.9 tok/s

    RTX 5090 32GB · unknown · NVFP4

    @tekizaihqGet

44. 

    Nemotron 3.5 Bolt 74B · 80 tok/s

    DGX Spark · llama.cpp · 2-bit GGUF + lilicorr drafter

    @filicrovalHF

45. 

    Qwen3.8-Flash-Next · 79.5 tok/s

    DGX Spark · EXL3 · EXL3 3.05 bpw

    @yume_arasaki

46. 

    Qwen3.8-Flash-Next · 79 tok/s

    RTX 3080 Ti 12GB · Strata · IQ2_XS

    @nigelinstantGet

47. 

    Qwen3.8-Flash-Next · 74.8 tok/s

    DGX Spark · TensorFold · NVFP4 MTP-6

    @redp314Get

48. 

    Qwen3.8-27B · 72 tok/s

    DGX Spark · unknown · NVFP4 + DFlash2

    @hasso5703Get

49. 

    Qwen3.8-Flash-Next · 72 tok/s

    RTX 3090 24GB · Strata · IQ3_S GSQ RCO GGUF

    @needmorevramGet

50. 

    Qwen3.8-Flash-Next · 70 tok/s

    RTX 3090 24GB · Strata · IQ3_S

    @superaleshaGet

51. 

    Qwen3.8-Flash-Next · 66.7 tok/s

    RTX 5070 12GB · Strata · IQ3_XXS GGUF

    @jun1228909Get

52. 

    AliceAI-Foundation-80B-A3B · 65 tok/s

    M5 Max 128GB · llama.cpp · Q4_K_M GGUF

    @aqtyHF

53. 

    Qwen3.8-Flash-Next · 65 tok/s

    DGX Spark · TensorFold · 4-bit QAD

    @sudoingX

54. 

    Qwen3.8-Flash-Next · 65 tok/s

    DGX Spark · NVIDIA · FP4-Local

    @jtdaviesHF

55. 

    Qwen3.8-27B · 61.9 tok/s

    M6 mini 32GB · TensorFold · MLX 4-bit + DFlash2

    @jmurillocodeHF

56. 

    GLM 5.3 Flash · 60.4 tok/s

    2× DGX Spark · TensorFold · EXL3 TR3 4bpw + DFlash2

    @MiaAI_labGet

57. 

    Qwen3.8-Flash-Next · 60 tok/s

    4× Tesla V100 32GB · Strata 0.1.43.1 · Q4_K_XL-DN4

    @DaSun64125381Get

58. 

    Qwen3.8-Flash-Next · 58 tok/s

    2× DGX Spark · vLLM · official FP8

    @sudoingX

59. 

    DeepSeek v4.1 Flash · 57.7 tok/s

    RTX PRO 6000 · llama.cpp b11412 · Q2

    @piyo621

60. 

    Qwen3.8-Flash-Next · 57 tok/s

    Tesla V100 32GB · Strata · Unsloth UD-IQ4_XS

    @tkysyskw

61. 

    GLM 5.3 Flash · 55.6 tok/s

    2x DGX Spark · unknown · NVFP4

    @PlusTen_AIGet

62. 

    Qwen3.8-Flash-Next · 55 tok/s

    RTX 4060 Ti 16GB · Strata 0.1.44.1 · IQ3_S

    @xiaobei1943Get

63. 

    Qwen3.8-Flash-Next · 55 tok/s

    RTX 4060 Ti 16GB · Strata · IQ3_S

    @xiaobei1943Get

64. 

    Ternary Bonsai 2 27B · 54.7 tok/s

    RTX 5060 Ti 16GB · llama.cpp · PTQ1_0

    @niklaslenz_aiHF

65. 

    CYBER-FROST 3.8 · 54.7 tok/s

    DGX Spark · ExLlamaV3 · EXL3 3.87 bpw

    @niklaslenz_ai

66. 

    Qwen3.8-Flash-Next · 53.8 tok/s

    Ryzen AI Max+ 395 128GB · Strata 0.1.40 · UD-IQ4_XS GGUF

    @Niko1221Get

67. 

    Qwen3.8-Flash-Next · 53.5 tok/s

    RTX 5070 Ti 16GB · Strata · IQ3_S GGUF

    @raminaol9ykasGet

68. U

    Underdog Saluki 27B · 53 tok/s

    DGX Spark · llama.cpp · 2-bit class

    @dedeneGet

69. 

    Qwen3.8-Flash-Next · 52.7 tok/s

    RTX 3080 10GB · Strata · IQ3_S GSQ RCO GGUF

    @needmorevramGet

70. 

    Qwen3.8-Flash-Next · 52 tok/s

    2× AMD MI50 32GB · Strata · IQ3_S

    @liucl

71. 

    Qwen3.8-Flash-Next Coder · 51 tok/s

    M5 Max 64GB · llama.cpp · ISTA GSQ-RCO

    @sebastavar

72. 

    Qwen3.8-Flash-Next · 50.2 tok/s

    DGX Spark · Vontra MLX · MLX 4-bit

    @openaiarka

73. 

    Empero Qwen3.8-35B-A3B · 50 tok/s

    RTX 3060 12GB · llama.cpp · Q4_K_M GGUF

    @Oluwaphilemon1

74. 

    Ternary Bonsai 2 27B · 50 tok/s

    RTX 3060 12GB · llama.cpp · Heretic GGUF

    @sudoingXHF

75. 

    Qwen3.8-Flash-Next · 50 tok/s

    RTX 5080 16GB · Strata · IQ3_S GGUF

    @rifaniponkGet

76. 

    Qwen3.8-Flash-Next · 50 tok/s

    RTX 4060 Ti · Strata · IQ3_S GGUF

    @xiaobei1943Get

77. 

    Qwen3.5-9B · 49.3 tok/s

    M4 Pro · rapid-mlx · 4-bit MLX

    @rapidmlx

78. 

    Qwen3-8B · 48.3 tok/s

    M4 Pro · rapid-mlx · 4-bit MLX

    @rapidmlx

79. 

    MiMo-V2.6-Distill-Qwen-9B · 47 tok/s

    RTX 5060 8GB · llama.cpp · Q5_K_M GGUF

    @stfu0911HF

80. 

    Qwen3.8-Flash-Next · 45 tok/s

    2× DGX Spark · unknown · FP8

    @Oluwaphilemon1

81. 

    Qwen3.8-27B · 42.3 tok/s

    DGX Spark · TensorFold 0.6.3 · NVFP4

    @edurdias

82. 

    Qwen3.8-Flash-Next · 42.1 tok/s

    RTX 5090 · crow-nest · unspecified

    @crowqntGet

83. 

    Qwen3.8-Flash-Next · 41.4 tok/s

    M5 Pro 64GB · Sushi · ~2.6 bits Sushi

    @aartiles24HF

84. 

    Qwen3.8-27B · 40.7 tok/s

    RTX 4090 24GB · llama.cpp · UD-Q4_K_XL GGUF

    @Oluwaphilemon1HF

85. 

    Ornith · 40 tok/s

    GTX 1660 SUPER 6GB · unknown · unlinked

    @mine_craft_buiHF

86. 

    Qwen3.8-Flash-Next · 40 tok/s

    2× RTX 3060 12GB · Strata · IQ3_S GGUF

    @xr_ymchnGet

87. 

    GLM 5.3 · 40 tok/s

    8× DGX Spark · vLLM · NVFP4

    @hypermac6502HF

88. 

    GLM 5.3 · 38.4 tok/s

    4x DGX Spark · TensorFold · EXL3 3.0bpw

    @niklaslenz_aiGet

89. 

    Qwen3.6-35B-A3B · 37 tok/s

    M1 Max 32GB · unknown · UD-IQ3_XXS GGUF

    @moonsteroidHF

90. 

    GLM 5.3 · 35 tok/s

    4× DGX Spark · TensorFold · NVFP4 + FP4 KV

    @majewskizby

91. 

    Qwen3.8-Flash-Next · 34.2 tok/s

    RTX 3090 + RTX 2070 SUPER · Strata · IQ3_S

    @OxtanedotxyzGet

92. R

    RED-SNOW 5.3 Flash · 33.0 tok/s

    DGX Spark · ExLlamaV3 · EXL3 2.49 bpw

    @niklaslenz_ai

93. 

    Qwen3.8-Flash-Next · 32.4 tok/s

    Intel Arc B570 10GB · XeStrata · Q2_0

    @llm_frontier_jp

94. 

    GLM 5.3 Flash · 32.3 tok/s

    DGX Spark · TensorFold · EXL3 2.05bpw

    @NFT_ChenHF

95. 

    Qwen3.8-Flash-Next · 32 tok/s

    DGX Spark · TensorFold · NVFP4

    @openaiarka

96. 

    Qwen3.8-27B · 31.8 tok/s

    DGX Spark · SGLang 0.5.21 · NVFP4

    @edurdias

97. 

    GLM 5.3 · 30 tok/s

    4x DGX Spark · sparkDash · Int4/Int8 mixed TP4

    @majewskizby

98. 

    GLM 5.3 Flash · 30 tok/s

    DGX Spark · TensorFold · EXL3 2.05bpw M288 mosaic 12L

    @filicroval

99. 

    Qwen3.8-27B · 28.7 tok/s

    DGX Spark · vLLM 0.30.0 · NVFP4

    @edurdias

100. 

     Qwen3.8-Flash-Next · 27.7 tok/s

     RTX 3060 12GB · Strata · IQ2_XS GGUF

     @dec21aiGet

101. 

     Qwen3.8-Flash-Next · 26.8 tok/s

     AMD MI50 32GB · Strata · IQ3_S

     @yokoyoko001

102. 

     Qwen3.8-27B · 23 tok/s

     RTX 5060 Ti 16GB · llama.cpp · GSQ-RCO IQ3_XXS-mtp

     @fntAInheadHF

103. 

     Qwen3.8-27B · 19 tok/s

     RTX 5060 Ti 16GB · llama.cpp · UD-Q2_K_XL

     @fntAInheadHF

104. 

     GLM 5.3 Flash · 18.6 tok/s

     DGX Spark · EXL3 · EXL3 K2

     @vcruz305Get

105. 

     GLM 5.3 Flash · 15 tok/s

     2x DGX Spark · vLLM · NVFP4

     @sudoingX

106. 

     Nex-N2.5-mini · 14 tok/s

     RTX 5060 Ti 16GB · llama.cpp · Q4_K_M GGUF

     @fntAInhead

107. 

     GLM 5.3 Flash · 11.3 tok/s

     Mac Studio M4 Max 64GB · unspecified · Q2

     @hokazuya

108. 

     Qwen3.8-27B TurboFCFusion · 10 tok/s

     RTX 5060 Ti 16GB · llama.cpp · IQ2_M GGUF

     @fntAInhead

109. 

     REAP50 GLM 5.3 Flash · 10 tok/s

     Radeon RX 9070 XT · Strata · Q2

     @syoyo

110. 

     Qwen3.8-Flash-Next · 9.0 tok/s

     Intel Arc B570 10GB · Strata · Q2_0

     @llm_frontier_jpGet

111. 

     Qwen3.8-Flash-Next · 7.8 tok/s

     M5 Max 64GB · LM Studio · GGUF Q4_K_M

     @miya_mos365

Models

• Qwen3.8-27B — Qwen · dense · 27B · new, hot · https://huggingface.co/Qwen/Qwen3.8-27B
• Qwen3.8-27B Mythos — medismera · dense · 27B · new, hot · https://huggingface.co/medismera/Qwen3.8-27B-OBLITERATED-Mythos-Class-Agentic
• Qwen3.8-35B-A3B APEX — IsValorum · MoE 35B-A3B · 35B / 3B act · new, hot · https://huggingface.co/IsValorum/Qwen3.8-35B-A3B-Distill-MTP-APEX-I-MiniPlus-V2.1-Abliterated-GGUF
• Huihui Qwen3.8-27B — huihui-ai · dense · 27B · new · https://huggingface.co/huihui-ai/Huihui-Qwen3.8-27B-abliterated
• Qwen3.8-27B OBLITERATED — OBLITERATUS · dense · 27B · new · https://huggingface.co/OBLITERATUS/Qwen3.8-27B-OBLITERATED
• Qwen3.8-27B Uncensored — orcarouter · dense · 27B · new · https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored
• Qwen3.8-27B Heretic — 0bserverx · dense · 27B · new · https://huggingface.co/0bserverx/Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF
• Qwen3.8-27B Splash — audreyt · dense · 27B · new · https://huggingface.co/audreyt/Qwen3.8-27B-Splash-abliterated
• SuperQwen3.8-27B — Jiunsong · dense · 27B · new · https://huggingface.co/Jiunsong/SuperQwen3.8-27b-abliterated
• Flash-Next RVN — 0bserverx · MoE · unknown · new · https://huggingface.co/0bserverx/RVN-Qwen3.8-Flash-Next-Abliterated-Uncensored
• Qwen3.8-27B TurboFC — community · dense · 27B · new
• Qwen3.8-27B Heretic ARA — trohrbaugh · dense · 27B · new · https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara
• Qwen3.8-27B Coletti — JonathanColetti · dense · 27B · new, hot · https://huggingface.co/JonathanColetti/Qwen3.8-27B-Uncensored
• Qwen3.8-27B NoRefusal — sss22213 · dense · 27B · new · https://huggingface.co/sss22213/Qwen3.8-27B-Heretic-NoRefusal
• Qwen3.8-27B KCRN — heterodoxin · dense · 27B · new · https://huggingface.co/heterodoxin/qwen-3.8-27b-abliterated
• Qwen3.8-27B Fable — DavidAU · dense · 27B · new · https://huggingface.co/DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU
• Qwen3.8-9B Heretic — Noobito45 · dense · 9B · new · https://huggingface.co/Noobito45/Qwen3.8-9B-heretic-uncensored-NVFP4-GGUF
• DeepSeek v4.1 Flash abliterated — distributedcognition · dense · unknown · new · https://huggingface.co/distributedcognition/DeepSeek-V4.1-Flash-abliterated
• Gemma-4 26B Abliterix — wangzhang · MoE · 26B-A4B · new · https://huggingface.co/wangzhang/gemma-4-26B-A4B-it-abliterix
• Gemma-4 26B Uncensored — TrevorJS · MoE · 26B-A4B · hot · https://huggingface.co/TrevorJS/gemma-4-26B-A4B-it-uncensored-GGUF
• Gemma-4 31B Uncensored — TrevorJS · dense · 31B · new · https://huggingface.co/TrevorJS/gemma-4-31B-it-uncensored-GGUF
• Gemma-4 12B OBLITERATED — OBLITERATUS · dense · 12B · new · https://huggingface.co/OBLITERATUS/Gemma-4-12B-OBLITERATED
• Qwen3.8-27B Ultra Heretic — llmfan46 · dense · 27B · new · https://huggingface.co/llmfan46/Qwen3.8-27B-Ultra-Uncensored-Heretic-Native-MTP-Preserved
• Qwen3.8-27B Aggressive — 0xKitkat · dense · 27B · new · https://huggingface.co/0xKitkat/Qwen3.8-27B-Uncensored-Aggressive
• Qwen3.8-27B Twin Turbo — DavidAU · dense · 27B · new, hot · https://huggingface.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored
• GLM 5.3 Flash abliterated — dealignai · MoE · unknown · new, hot · https://huggingface.co/dealignai/GLM-5.3-Flash-ABLITERATED-FP8
• GLM 5.3 Flash Uncensored — orcarouter · MoE · 320B / 18B act · new · https://huggingface.co/orcarouter/GLM-5.3-Flash-Uncensored-GGUF
• GLM 5.3 Uncensored — dealignai · MoE · 753B · new · https://huggingface.co/dealignai/GLM-5.3-UNCENSORED-FP8
• GLM 5.3 EXL3 abliterated — drowzeys · MoE · unknown · new · https://huggingface.co/drowzeys/keys-GLM-5.3-EXL3-Abliterated
• Qwen3.6-35B-A3B — Qwen · MoE 35B-A3B · 35B / 3B act · https://huggingface.co/Qwen/Qwen3.6-35B-A3B
• Qwen3.6-35B-A3B abliterated — community · MoE 35B-A3B · 35B / 3B act · hot
• Empero Qwen3.8-35B-A3B — empero-ai · MoE distill · 35B / 3B act · new, hot
• Qwen3.8-Flash-Next — Qwen · dense · unknown · new, hot
• Qwen3.5 2B — Qwen · dense · 2B · new, hot
• Qwen3.5-4B — Qwen · dense · 4B · new
• Qwen3.5-9B — Qwen · dense · 9B · new
• Qwen3-8B — Qwen · dense · 8B
• Gemma-4 26B-A4B-it — Google / community · MoE · 26B-A4B · new, hot
• Nex-N2.5-mini — community · MoE post-train · 35B-A3B class · new
• Ornith — ornith-ai · unknown · unknown · new · https://huggingface.co/ornith-ai
• GLM 5.3 Flash — Zhipu · dense · unknown · new, hot
• DeepSeek v4.1 Flash — DeepSeek · dense · unknown · new, hot
• MiMo-V2.6-Distill-Qwen-9B — Xiaomi MiMo · dense · 9B · new, hot
• AliceAI-Foundation-80B-A3B — Yandex · MoE 80B-A3B · 80B / 3B act · new · https://huggingface.co/Yamada114514/AliceAI-Foundation-80B-A3B-Base-GGUF
• MiMo-V2.6-Flash — Xiaomi MiMo · MoE · unknown · new
• Ling-3.0-Flash — inclusionAI · dense · unknown · new · https://huggingface.co/inclusionAI/Ling-3.0-flash-int4
• Nemotron 3.5 Lightning — NVIDIA · MoE 30B-A3B · 30B / 3B act · new
• Nemotron 3.5 Bolt 74B — NVIDIA · MoE 74B · 74B · new · https://huggingface.co/nvidia/Nemotron-3.5-Bolt
• GLM-5.2 — Zhipu · MoE · unknown · new
• Kimi K3 — Moonshot · MoE · 2.8T / 104B act · new, hot · https://huggingface.co/moonshotai/Kimi-K3
• CYBER-FROST 3.8 — Blackfrost-AI · MoE · ~180B · new · https://huggingface.co/Blackfrost-AI/CYBER-FROST-3.8-BF16
• Qwen-Image-2.1 — Qwen · image · 7B visual · new, hot · https://huggingface.co/Qwen/Qwen-Image-2.1
• Qwen-Image-2.1 Uncensored — abenzerps · image · GGUF · https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-GGUF
• Ternary Bonsai 2 27B — PrismML · dense ternary · 27B · new, hot · https://huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf
• MiniMax-M2.5 — MiniMax · unknown · unknown · new
• Swift 1.5 Qwen3.8 Flash-Next — UkisAI · dense · 27B-class · new, hot
• GLM 5.3 — Zhipu · MoE · 753B · new
• Qwen3.8-Flash-Next Coder — ISTA-DASLab · MoE prune · Flash-Next, 256/512 experts · new
• REAP50 GLM 5.3 Flash — community · MoE prune · REAP50 of Flash · new
• Underdog Saluki 27B — Underdog AI · dense · 27B · new
• Gemma-4 26B-A4B-it — Google · MoE · 26B-A4B · new · https://huggingface.co/google/gemma-4-26b-a4b-it
• RED-SNOW 5.3 Flash — Blackfrost-AI · MoE · GLM-5.3-Flash class · new

Hardware

• M4 Pro — Apple · 24–64 GB unified · 4 pairs
• M5 Max — Apple · 64–128 GB unified · 4 pairs
• DGX Spark — NVIDIA · 128 GB unified · 33 pairs
• RTX 5090 — NVIDIA · 32 GB VRAM · 12 pairs
• RTX 4090 — NVIDIA · 24 GB VRAM · 3 pairs
• RTX 5060 Ti — NVIDIA · 8–16 GB VRAM · 5 pairs
• RTX 5060 — NVIDIA · 8 GB VRAM · 1 pairs
• RTX 3060 — NVIDIA · 8–12 GB VRAM · 3 pairs
• GTX 1660 SUPER — NVIDIA · 6 GB VRAM · 1 pairs
• Tesla V100 32GB — NVIDIA · 32 GB VRAM · 1 pairs
• 2x Tesla V100 32GB — NVIDIA · 64 GB VRAM (2x32) · 0 pairs
• 4× Tesla V100 32GB — NVIDIA · 128 GB VRAM (4×32) · 1 pairs
• M4 Max 64GB — Apple · 64 GB unified · 1 pairs
• M1 Max 64GB — Apple · 64 GB unified · 0 pairs
• M1 Max 32GB — Apple · 32 GB unified · 1 pairs
• M4 mini 24GB — Apple · 24 GB unified · 0 pairs
• M3 Ultra — Apple · 96-512 GB unified · 0 pairs
• RTX 5070 12GB — NVIDIA · 12 GB VRAM · 1 pairs
• RTX 3090 24GB — NVIDIA · 24 GB VRAM · 5 pairs
• 2x RTX 3090 24GB — NVIDIA · 48 GB VRAM (2x24) · 3 pairs
• M5 Pro 64GB — Apple · 64 GB unified · 2 pairs
• M6 mini 32GB — Apple · 32 GB unified · 1 pairs
• RTX A5000 — NVIDIA · 24 GB VRAM · 2 pairs
• RTX 3080 10GB — NVIDIA · 10 GB VRAM · 1 pairs
• RTX 3080 Ti 12GB — NVIDIA · 12 GB VRAM · 1 pairs
• RTX 4060 Ti — NVIDIA · 8–16 GB VRAM, size not in the quote · 3 pairs
• RTX PRO 6000 + 2× GB10 — NVIDIA · PRO 6000 + 2× GB10, sizes not in the quote · 1 pairs
• 4× RTX PRO 6000 — NVIDIA · 4× PRO 6000, 250 W each · 1 pairs
• 2× RTX PRO 6000 — NVIDIA · 2× PRO 6000, 300 W each · 1 pairs
• RTX PRO 6000 Blackwell Max-Q — NVIDIA · PRO 6000 Max-Q, 250W, VRAM size not in the quote · 2 pairs
• RTX PRO 6000 — NVIDIA · 96 GB class, 81–85 GB used in the quote · 2 pairs
• M4 Pro 64GB — Apple · 64 GB unified · 1 pairs
• 2× RTX 5060 Ti 16GB — NVIDIA · 32 GB VRAM (2×16) · 1 pairs
• RTX 5070 Ti 16GB — NVIDIA · 16 GB VRAM · 1 pairs
• RTX 5080 16GB — NVIDIA · 16 GB VRAM · 1 pairs
• 2× RTX 3060 12GB — NVIDIA · 24 GB VRAM (2×12) · 1 pairs
• Ryzen AI Max+ 395 128GB — AMD · 128 GB unified · 1 pairs
• AMD MI50 32GB — AMD · 32 GB VRAM · 1 pairs
• 2× AMD MI50 32GB — AMD · 64 GB VRAM (2×32) · 1 pairs
• Radeon RX 9070 XT — AMD · 16 GB VRAM + 64 GB DDR4 host · 1 pairs
• 2× DGX Station GB300 — NVIDIA · 2× GB300, size not in the quote · 1 pairs
• Intel Arc B570 10GB — Intel · 10 GB VRAM · 2 pairs
• Radeon RX 9060 XT 16GB — AMD · 16 GB VRAM · 1 pairs
• RTX 3090 + RTX 2070 SUPER — NVIDIA · mixed VRAM, sizes not in the quote · 1 pairs
• RTX 5060 Ti 16GB + RTX 3070 8GB — NVIDIA · 24 GB VRAM (16+8) · 1 pairs

Tape

• 2026-10-18 @crowqnt — Qwen3.8-Flash-Next crow-nest on RTX 5090: 42.1 avg decode after improvements. https://x.com/crowqnt/status/2111771587301298388
• 2026-10-18 @dextune — Flash-Next UD-IQ4_XS Strata on RTX 4090 + 64GB: 92 decode at 136K needle. https://x.com/dextune/status/2111733185084350539
• 2026-10-18 @Tech2Wild — DeepSeek-V4.1-Flash TensorFold on 2× Spark (2nd cable): 93.6 one-stream. https://x.com/Tech2Wild/status/2111666374032425064
• 2026-10-18 @jtdavies — NVIDIA Qwen3.8-Flash-Next FP4-Local on DGX Spark at 1M YaRN: 65 tok/s. https://x.com/jtdavies/status/2111593677684687152
• 2026-09-27 @jmurillocode — Qwen3.8-27B TensorFold MLX 4-bit DFlash2 on 32GB Mac mini M6: 61.9 code / 30.8 chat. https://x.com/jmurillocode/status/2104138445928956021
• 2026-09-27 @aartiles24 — Qwen3.8-27B TensorFold MLX 4-bit DFlash2 on MacBook M5 Pro 64GB: 89.9 tok/s. https://x.com/aartiles24/status/2104179801388884125
• 2026-09-27 @Oluwaphilemon1 — Qwen3.8-27B TensorFold DFlash2 on one Spark: 102.9 tok/s. https://x.com/Oluwaphilemon1/status/2104031407194427892
• 2026-09-28 @PlusTen_AI — GLM 5.3 Flash NVFP4 on 2x DGX Spark: 55.6 coding / 33.6 Korean prose. https://x.com/PlusTen_AI/status/2104531371146482165
• 2026-09-28 @DaSun64125381 — Flash-Next IQ3_S Strata on RTX 5090 256K: 104 decode / 1657 prefill. https://x.com/DaSun64125381/status/2104536066028192175
• 2026-09-29 @moonsteroid — Qwen3.6-35B-A3B UD-IQ3_XXS on MBP M1 Max 32GB: 37 decode / 350 prefill at 100k. https://x.com/moonsteroid/status/2104903466661314585
• 2026-09-29 @needmorevram — Flash-Next IQ3_S GSQ RCO Strata on one RTX 3090 at 128K: 72 decode. https://x.com/needmorevram/status/2104898095318544540
• 2026-09-30 @sudoingX — GLM 5.3 Flash NVFP4 stock vLLM on 2x Spark: 15 decode MTP-off. https://x.com/sudoingX/status/2105246609730924695
• 2026-09-30 @dec21ai — Flash-Next IQ2_XS Strata on RTX 3060 12GB Thinking Medium: 27.7 decode. https://x.com/dec21ai/status/2105260285439422767
• 2026-09-30 @Knuckles_XBT — Swift 1.5 Flash-Next IQ2_XS Strata on RTX 4090 24GB at 256K: ~220 decode / ~3830 prefill. https://x.com/Knuckles_XBT/status/2105245662657011841
• 2026-10-01 @Ja6ek — Qwen3.8-27B NVFP4 on RTX A5000 at 262K: TensorFold 177.57 / vLLM 139.31. https://x.com/Ja6ek/status/2105623160649298395
• 2026-10-01 @arianpg — Flash-Next IQ3_S Strata on RTX 5090 + 128GB RAM: 142 decode. https://x.com/arianpg/status/2105625971261161895
• 2026-10-01 @redp314 — Flash-Next NVFP4 MTP-6 TensorFold 0.3.6.3 on one DGX Spark: 74.8 single-stream. https://x.com/redp314/status/2105569144380739979
• 2026-10-02 @draslan_eth — Flash-Next IQ2_XS Strata 0.1.31 on one RTX 3090 at 32K: 100.6 decode. https://x.com/draslan_eth/status/2105946305646502091
• 2026-10-02 @majewskizby — Full GLM 5.3 753B Int4/Int8 TP4 on 4× DGX Spark: 30.0 prose decode, thinking off. https://x.com/majewskizby/status/2105962150158041477
• 2026-10-02 @needmorevram — Flash-Next IQ3_S GSQ RCO Strata on RTX 3080 10GB at 128K: 52.7 decode. https://x.com/needmorevram/status/2106081480321732643
• 2026-10-03 @xiaobei1943 — Flash-Next IQ3_S Strata on RTX 4060 Ti at 128K: 50 decode average. https://x.com/xiaobei1943/status/2106351673845350725
• 2026-10-03 @hokazuya — GLM 5.3 Flash EXL3 4bpw TensorFold on RTX PRO 6000 + 2× GB10: 120 decode, thinking off. https://x.com/hokazuya/status/2106341022934724999
• 2026-10-04 @llm_frontier_jp — Qwen3.8-27B NVFP4 on RTX PRO 6000 Max-Q: TensorFold 272.8 / vLLM 151.4 code greedy. https://x.com/llm_frontier_jp/status/2106575030448800001
• 2026-10-04 @draslan_eth — Qwen3.6-35B-A3B 4-bit Rapid-MLX on M4 Pro 64GB: 108.6 decode. https://x.com/draslan_eth/status/2106705583718154587
• 2026-10-03 @Kurcide — GLM 5.3 Flash EXL3 TensorFold on 4× RTX PRO 6000: 264 prose one-stream. https://x.com/Kurcide/status/2106460933132595499
• 2026-10-04 @edurdias — Qwen3.8-27B NVFP4 on one GB10: TensorFold 42.3 / vLLM 28.7 / SGLang 31.8. https://x.com/edurdias/status/2106587941472440463
• 2026-10-05 @Kurcide — GLM 5.3 Flash EXL3 TensorFold on 2× RTX PRO 6000: 238 prose one-stream. https://x.com/Kurcide/status/2106846688606331298
• 2026-10-04 @MehdiHacks — GLM 5.3 Flash EXL3 TensorFold on 2× DGX Spark: 85.0 code, one stream. https://x.com/MehdiHacks/status/2106815809700135038
• 2026-10-05 @needmorevram — Flash-Next IQ3_S GSQ RCO Strata on 2× RTX 5060 Ti 16GB: 99.9 completed-request decode. https://x.com/needmorevram/status/2106965241850646980
• 2026-10-04 @xr_ymchn — Flash-Next official IQ3_S Strata on 2× RTX 3060 12GB: 40 short-context decode. https://x.com/xr_ymchn/status/2106790441903681570
• 2026-10-05 @raminaol9ykas — Flash-Next IQ3_S Strata on RTX 5070 Ti 16GB: 53.5 cached decode. https://x.com/raminaol9ykas/status/2107043129626955897
• 2026-10-05 @jun1228909 — Flash-Next IQ3_XXS Strata on RTX 5070: 66.7 decode after PCIe 3.0 x16. https://x.com/jun1228909/status/2107036816071422172
• 2026-10-05 @miya_mos365 — Flash-Next Q4_K_M LM Studio CPU-only on M5 Max 64GB: 7.80 tok/s, thinking on. https://x.com/miya_mos365/status/2107046489965084969
• 2026-10-05 @piyo621 — Flash-Next Q4_XL llama.cpp b11412 on RTX PRO 6000: 92.82 t/s. DeepSeek-V4-Flash Q2: 57.70. https://x.com/piyo621/status/2107085400473010303
• 2026-10-06 @sebastavar — Flash-Next Coder ISTA GSQ-RCO llama.cpp on M5 Max 64GB: 51 writing. https://x.com/sebastavar/status/2107334963343331608
• 2026-10-06 @NFT_Chen — GLM 5.3 Flash EXL3 2.05bpw TensorFold on one Spark: 32.32 decode. https://x.com/NFT_Chen/status/2107342944072241520
• 2026-10-06 @niklaslenz_ai — Full GLM 5.3 EXL3 3.0bpw TensorFold on 4× Spark: 38.4 code / 24.9 prose. https://x.com/niklaslenz_ai/status/2107391906963411430
• 2026-10-06 @Niko1221 — Flash-Next UD-IQ4_XS Strata 0.1.40 on Ryzen AI Max+ 395 128GB: 53.8 decode / 1,293 prefill at 8K. https://github.com/Niko1221/Strata/releases/tag/v0.1.40
• 2026-10-07 @yokoyoko001 — Flash-Next IQ3_S Strata on one MI50 32GB + 70GB DDR3: 26.82 decode. https://x.com/yokoyoko001/status/2107768586970571262
• 2026-10-07 @liucl — Flash-Next IQ3_S Strata on 2× MI50 32GB: 52 decode / 645 prefill. https://x.com/liucl/status/2107740068194722155
• 2026-10-07 @openaiarka — Flash-Next on one Spark: TensorFold NVFP4 32 decode, Vontra MLX 4-bit 50.16. https://x.com/openaiarka/status/2107728765480812607
• 2026-10-07 @sudoingX — Flash-Next 4-bit QAD TensorFold on one Spark: 65 code. Official FP8 vLLM on 2× Spark: 58. https://x.com/sudoingX/status/2107789169322082730
• 2026-10-07 @majewskizby — Full GLM 5.3 NVFP4 + FP4 KV TensorFold on 4× Spark at 256K: 35.0 prose / 42.0 code. https://x.com/majewskizby/status/2107711362289828332
• 2026-10-07 @hypermac6502 — GLM 5.3 NVFP4 vLLM on 8× DGX Spark: C1 40 tok/s. https://x.com/hypermac6502/status/2107915407277621484
• 2026-10-08 @SJRonanMD — Fused GLM 5.3 NVFP4 SGLang on 2× DGX Station GB300: 81.5 single stream. https://x.com/SJRonanMD/status/2108075612049150061
• 2026-10-08 @syoyo — REAP50 GLM 5.3 Flash Q2 Strata on Radeon RX 9070 XT: 10 decode / 200 prefill. https://x.com/syoyo/status/2108162708843671972
• 2026-10-08 @filicroval — GLM 5.3 Flash EXL3 2.05bpw M288 mosaic 12L TensorFold on one GB10: 30 tok/s at 262K. https://x.com/filicroval/status/2108164968247947316
• 2026-10-08 @geldeki — Flash-Next IQ3_S GSQ-RCO Strata on 2× RTX 3090: 122.5 peak decode, 2,100 prefill in the logs. https://x.com/geldeki/status/2108149127850639598
• 2026-10-08 @tkysyskw — Flash-Next UD-IQ4_XS Strata on Tesla V100 32GB at 131K: 57 decode. https://x.com/tkysyskw/status/2108041292076974307
• 2026-10-09 @rifaniponk — Flash-Next IQ3_S Strata on RTX 5080, vision on: 50 decode. Prefill range not ranked. https://x.com/rifaniponk/status/2108382777557315771
• 2026-10-09 @niklaslenz_ai — RED-SNOW 5.3 Flash EXL3 2.49 bpw on one Spark: 32.95 median long outputs. CYBER-FROST 3.8 EXL3 3.87 bpw: 54.7 code. https://x.com/niklaslenz_ai/status/2108524496621539492
• 2026-10-10 @orenoa7 — Flash-Next Q2_0 GSQ-RCO Strata on RTX 5090: 191 JP prose at 4K. https://x.com/orenoa7/status/2108756820403843081
• 2026-10-10 @dedene — Underdog Saluki 27B + DFlash2 on one DGX Spark: 53 tok/s (from 24). https://x.com/dedene/status/2108879473726910650
• 2026-10-11 @RubenTorney — Flash-Next IQ3_S Strata on one RTX 3090: 100 tok/s short after patches (99 at 56K, 82 at 256K). https://x.com/RubenTorney/status/2109088323515879622
• 2026-10-11 @therealskrlr — Flash-Next IQ3_S Strata on RTX 5090 + 64GB: 195.8 tok/s generation / 3,944 prompt for a long one-shot. https://x.com/therealskrlr/status/2109129375446958271
• 2026-10-11 @geldeki — Flash-Next IQ3_S Strata v0.1.42 on 2× RTX 3090: peak 153.4 decode, cold prefill 2,844 at 191K. https://x.com/geldeki/status/2109208295957201171
• 2026-10-12 @Tech2Wild — GLM 5.3 Flash TensorFold on 4× DGX Spark: 127 code (vs 120 vLLM). https://x.com/Tech2Wild/status/2109456865859928528
• 2026-10-12 @0xSero — GLM 5.3 Flash EXL3 4 bpw on 4× DGX Spark: 104 prose / 152 code. https://x.com/0xSero/status/2109620260282188273
• 2026-10-12 @geldeki — Flash-Next IQ3_S Strata v0.1.43 on 2× RTX 3090: peak 158.3 decode / 3,149 prefill. https://x.com/geldeki/status/2109610268938408179
• 2026-10-11 @superalesha — Flash-Next on one RTX 3090 Strata: IQ3_S 70 decode (66 at 261K), IQ2_XS 87 decode. https://x.com/superalesha/status/2109422253364375946
• 2026-10-12 @nigelinstant — Flash-Next IQ2_XS Strata v0.1.43 on RTX 3080 Ti 12GB: 79.0 decode at 101K (short range 93–126 not ranked). https://x.com/nigelinstant/status/2109583943385460837
• 2026-10-12 @mizuirot — Flash-Next UD-Q4_K_XL on modified llama.cpp, single RTX 5090: 114.25 decode / 2,141.16 prefill. https://x.com/mizuirot/status/2109608382961512689
• 2026-10-13 @DaSun64125381 — Flash-Next Q4_K_XL-DN4 Strata 0.1.43.1 on 4× V100 32GB: 60 single / 204 aggregate. https://x.com/DaSun64125381/status/2109976522375520296
• 2026-10-13 @hokazuya — GLM 5.3 Flash Q2 on Mac Studio M4 Max 64GB: 11.33 tok/s. https://x.com/hokazuya/status/2109980269508018566
• 2026-10-14 @Oxtanedotxyz — Flash-Next IQ3_S Strata on 3090 + 2070 SUPER: 34.2. https://x.com/Oxtanedotxyz/status/2110333416956907772
• 2026-10-14 @RadxaNaoki — Gemma-4 26B-A4B UD-Q4_K_XL llama.cpp v0.6.0 on RX 9060 XT 16GB: 88 decode / 2464 prefill. https://x.com/RadxaNaoki/status/2110323193726861634
• 2026-10-14 @llm_frontier_jp — Flash-Next Q2_0 on Arc B570 10GB: Strata 9.02 / XeStrata 32.4 with MTP. https://x.com/llm_frontier_jp/status/2110331888716153091
• 2026-10-13 @sudoingX — DeepSeek v4.1 Flash custom vLLM on 2× DGX Spark: 103.0 code / 57.9 prose / 141.0 structured. https://x.com/sudoingX/status/2109957456709472609
• 2026-10-15 @sudoingX — Ternary Bonsai 2 Heretic on RTX 3060 12GB: 50 tok/s with draft head (39 without), 10.7 GB. https://x.com/sudoingX/status/2110451458302943383
• 2026-10-15 @VectorCrossProd — Flash-Next Coder IQ1_M Strata v0.1.44 on RTX 5060 Ti 16GB + RTX 3070 8GB: 127 decode at 154K (peak 154). https://x.com/VectorCrossProd/status/2110643045985366075
• 2026-10-15 @niklaslenz_ai — Ternary Bonsai 2 27B PTQ1_0 on RTX 5060 Ti 16GB: 54.7 decode thinking off (IQ4_XS companion 26.6). https://x.com/niklaslenz_ai/status/2110702234950676641
• 2026-10-15 @aartiles24 — Qwen 3.8 Flash Next ~2.6 bits Sushi on MacBook Pro M5 Pro 64GB: 41.4 decode in a 17 h agent run. https://x.com/aartiles24/status/2110638559333949567
• 2026-10-17 @filicroval — Nemotron 3.5 Bolt 74B 2-bit GGUF + lilicorr on llama.cpp, one DGX Spark: 80 tok/s (math 108 / code 81 / prose 63). https://x.com/filicroval/status/2111401907830849845
• 2026-10-14 @Oluwaphilemon1 — Qwen3.8-27B on RTX 5090: TensorFold 0.6.6 + DFlash2 370 tok/s code; ExLlamaV3 258; llama.cpp MTP IQ4_XS 186. https://x.com/Oluwaphilemon1/status/2110224268222992606
• 2026-10-14 @xiaobei1943 — Flash-Next IQ3_S Strata v0.1.44.1 on RTX 4060 Ti 16GB: stable average 55 tok/s, 262K. https://x.com/xiaobei1943/status/2110312405435830666
• 2026-09-23 @hasso5703 — Qwen3.8-27B on one Spark: SGLang, NVFP4, DFlash2, 72 tok/s greedy median. Repo claim. https://github.com/hasso5703/dgx-spark-qwen38
• 2026-09-10 @vcruz305 — GLM-5.3-Flash EXL3 K2 on one Spark: SGLang, MTP k=2, 18.57 tok/s. Repo claim. https://github.com/vcruz305/GLM-5.3-Flash-EXL3-K2-SGLang-DGX-Spark-recipe
• 2026-09-22 @stfu0911 — MiMo-V2.6 Distill Qwen-9B Q5_K_M on RTX 5060 8GB: ~47 decode, ~1600 prefill, 262k. https://x.com/stfu0911/status/2102409155143409688
• 2026-09-22 @aqty — AliceAI 80B-A3B Q4_K_M on M5 Max 128GB: ~65 tok/s, 45.1 GiB. https://x.com/aqty/status/2102413529206976799
• 2026-09-21 @tekizaihq — Flash-Next NVFP4 on one 5090: 80.9 tok/s single-stream. https://x.com/tekizaihq/status/2101844907618939233
• 2026-09-20 @yume_arasaki — Flash-Next EXL3 on one Spark: 79.5 code reproduced; 102.6 repetitive clamps; 71.7 prose. https://x.com/yume_arasaki/status/2101741448811229219
• 2026-09-20 @ViC305 — EXL3 Spark recipe reproduced at 79.5 on Yume’s box. https://x.com/ViC305/status/2101747348322103706
• 2026-09-30 @MiaAI_lab — GLM-5.3-Flash EXL3 on 2× DGX Spark with TensorFold: 60.4 tok/s prose, 114.7 structured single-stream; 4 streams 108.8 / 227.9 aggregate. Repo claim. https://github.com/MiaAI-Lab/GLM-5.3-Flash-EXL3-2x-DGX-Sparks-TensorFold
• 2026-09-20 @MiaAI_lab — Spark street: Flash-Next solo; GLM 5.3 / DeepSeek v4.1 Flash on 2×+; orch+worker+scanner at 6×. https://x.com/MiaAI_lab/status/2101821681404624976
• 2026-09-21 @sethforprivacy — 8 Sparks: 4× GLM ring, 2× Flash-Next, 2× DeepSeek v4f. https://x.com/sethforprivacy/status/2101834433087070708
• 2026-09-17 @fillagrew — 5090 bake: 2B at 351 tok/s, abliterated 26B-A4B at 173. https://x.com/fillagrew/status/2100478211134025937
• 2026-09-17 @fntAInhead — 5060 Ti 16GB llama.cpp bake-off: 23 / 19 / 14 / 10 tok/s. https://x.com/fntAInhead/status/2100499089431359808
• 2026-09-17 @Youssofal_ — Flash-Next on M5 Max via MTPLX: 126.5 peak, 50 at 200k. https://x.com/Youssofal_/status/2100468205030719533
• 2026-09-17 @Oluwaphilemon1 — Official FP8 Flash-Next on 2× Spark, ~45 tok/s sustained. https://x.com/Oluwaphilemon1/status/2100498960645308573
• 2026-09-17 @Oluwaphilemon1 — Qwen3.8-27B UD-Q4_K_XL on 4090: 40.7 decode, 260k-class. https://x.com/Oluwaphilemon1/status/2100409183115821394
• 2026-09-17 @bonellisystems — Abliterated 35B-A3B NVFP4 on 2× Spark: 94.4 decode. https://x.com/bonellisystems/status/2100417191434768450
• 2026-09-17 @mine_craft_bui — Ornith ~40–46 tok/s on a 1660 SUPER 6GB. https://x.com/mine_craft_bui/status/2100492671534166056
• 2026-09-17 @Oluwaphilemon1 — Empero 35B-A3B Q4_K_M on a 12GB 3060: ~50 decode. https://x.com/Oluwaphilemon1/status/2100374395482939401
• 2026-09-16 @rapidmlx — Locked suite on one M4 Pro: 85.5 tok/s at 21 GB peak. https://x.com/rapidmlx/status/2100252655990010209

Street stacks

Recipes on DGX Spark. Not tok/s pairs. Source @MiaAI_lab.

• 1× Solo — qwen3.8-flash-next (Solo load); qwen3.8-27b (27B, SGLang); ling-3.0-flash; glm-5.3-flash (EXL3 K2, one box)
• 2× Dual — glm-5.3-flash (GLM on both); glm-5.3-flash (TensorFold, 4 streams); deepseek-v4.1-flash; qwen3.8-flash-next; mimo-v2.6-flash
• 3× Triple — glm-5.3-flash + qwen3.8-flash-next (GLM orch on 2×, Qwen worker on 1×); glm-5.3-flash (All three); deepseek-v4.1-flash (Homogeneous); glm-5.2 (NVFP4)
• 4× Quad — glm-5.3-flash + qwen3.8-flash-next (GLM orch 2× + Qwen worker 2×); glm-5.3-flash (Homogeneous); deepseek-v4.1-flash (Homogeneous)
• 6× 6+ — glm-5.3-flash + qwen3.8-flash-next + deepseek-v4.1-flash (GLM orch + Qwen worker + DeepSeek scanner)

Back to Home · Manifesto · Terms