Logic and Math · Latency

Latency on Logic and Math

Logic and Math pooled: 3,288 questions; only setups that finished both. How to read this page · Method

Latency

Milliseconds from request to reply per question (log scale): the median (family colour) and the 90th percentile (grey, nearest rank); beside them questions per minute, wall clock from the run's first to last answer, and requests in flight on average over its active time · lower latency is better

  1. GLiDE 0.4B

    GLiDE 0.4B · Fastino · decision model

    Parameters: 0.40B dense

    Latency: median 11 ms · p90 14 ms

    Logic 23.2% · Math 20.6%

    Its run record (Method) →

    11 ms14 ms4,286.1≈ 0.8
  2. GLiClass v4 Edge

    GLiClass v4 Edge · decision model

    Parameters: 0.4B dense

    Base model: DeBERTa-v3-large

    Latency: median 12 ms · p90 23 ms

    Logic 23.6% · Math 22.7%

    Its run record (Method) →

    12 ms23 ms3,520.7≈ 0.9
  3. OneJev1.6-0.8B

    OneJev1.6-0.8B · OmniJev · decision model

    Parameters: 0.8B dense

    Latency: median 15 ms · p90 17 ms

    Logic 27.0% · Math 28.2%

    Its run record (Method) →

    15 ms17 ms3,180.0≈ 0.8
  4. d1-2B

    d1-2B · Liquid AI · decision model

    Parameters: 1.92B dense

    Base model: LFM3-VL-2B

    Latency: median 17 ms · p90 21 ms

    Logic 30.7% · Math 37.1%

    Its run record (Method) →

    17 ms21 ms2,899.4≈ 0.8
  5. GLiDE 0.8B

    GLiDE 0.8B · Fastino · decision model

    Parameters: 0.75B dense

    Latency: median 20 ms · p90 25 ms

    Logic 29.2% · Math 32.1%

    Its run record (Method) →

    20 ms25 ms2,560.5≈ 0.9
  6. StartLux-Decision-v1.1-0.8B-Instant

    StartLux-Decision-v1.1-0.8B-Instant · StartLux · decision model

    Parameters: 0.8B dense

    Base model: Qwen3.5-0.8B

    Latency: median 24 ms · p90 29 ms

    Logic 24.7% · Math 29.7%

    Its run record (Method) →

    24 ms29 ms2,190.7≈ 0.9
  7. KnowLine-0.8B-Gen1

    KnowLine-0.8B-Gen1 · PelaAI · decision model

    Parameters: 0.8B dense

    Base model: Qwen3.5-0.8B

    Latency: median 26 ms · p90 35 ms

    Logic 26.7% · Math 28.2%

    Its run record (Method) →

    26 ms35 ms1,971.6≈ 0.9
  8. GLiDE 2B

    GLiDE 2B · Fastino · decision model

    Parameters: 1.88B dense

    Latency: median 28 ms · p90 37 ms

    Logic 32.3% · Math 36.8%

    Its run record (Method) →

    28 ms37 ms1,825.6≈ 0.9
  9. Perplexity Decider Flash v1.3

    Perplexity Decider Flash v1.3 · Perplexity · decision model

    Parameters: 4B dense

    Latency: median 29 ms · p90 35 ms

    Logic 44.7% · Math 55.9%

    Its run record (Method) →

    29 ms35 ms1,744.8≈ 0.9
  10. Chinese Decision Pointer 4B

    Chinese Decision Pointer 4B · quincyqiang · decision model

    Parameters: 4B dense

    Base model: Qwen3.5-4B

    Latency: median 30 ms · p90 42 ms

    Logic 40.4% · Math 50.4%

    Its run record (Method) →

    30 ms42 ms1,656.8≈ 0.9
  11. vera-medium-omni

    vera-medium-omni · scar-ai · decision model

    Parameters: 4.2B dense

    Base model: Qwen3.5-4B

    Latency: median 31 ms · p90 36 ms

    Logic 44.1% · Math 56.8%

    Its run record (Method) →

    31 ms36 ms1,714.4≈ 0.9
  12. Strands Decider 2B

    Strands Decider 2B · Strands Agents · decision model

    Parameters: 2B dense

    Latency: median 35 ms · p90 42 ms

    Logic 22.4% · Math 28.7%

    Its run record (Method) →

    35 ms42 ms1,552.4≈ 0.9
  13. StartLux-Decision-v1.1-2B-Instant

    StartLux-Decision-v1.1-2B-Instant · StartLux · decision model

    Parameters: 2B dense

    Latency: median 36 ms · p90 42 ms

    Logic 30.8% · Math 38.7%

    Its run record (Method) →

    36 ms42 ms1,505.0≈ 0.9
  14. Cygnet

    Cygnet · Blockbrain Labs · decision model

    Parameters: 11.95B dense

    Base model: Gemma 4 12B IT

    Latency: median 38 ms · p90 44 ms

    Logic 39.3% · Math 44.3%

    Its run record (Method) →

    38 ms44 ms1,449.7≈ 0.9
  15. Adjudex 9B v5

    Adjudex 9B v5 · newsbang · decision model

    Parameters: 9B dense

    Base model: Qwen3.5-9B

    Latency: median 38 ms · p90 45 ms

    Logic 46.4% · Math 62.1%

    Its run record (Method) →

    38 ms45 ms1,471.3≈ 0.9
  16. KnowLine-9B-Gen1

    KnowLine-9B-Gen1 · PelaAI · decision model

    Parameters: 9B dense

    Base model: Qwen3.5-9B

    Latency: median 38 ms · p90 55 ms

    Logic 44.3% · Math 59.8%

    Its run record (Method) →

    38 ms55 ms1,388.5≈ 1.0
  17. Jeffan-9B-1211

    Jeffan-9B-1211 · tattrongvu · decision model

    Parameters: 9B dense

    Base model: Qwen3.5-9B

    Latency: median 39 ms · p90 53 ms

    Logic 48.2% · Math 59.8%

    Its run record (Method) →

    39 ms53 ms1,280.3≈ 0.9
  18. Perplexity Decider Flash v1.2

    Perplexity Decider Flash v1.2 · Perplexity · decision model

    Parameters: 4B dense

    Latency: median 41 ms · p90 51 ms

    Logic 43.1% · Math 55.2%

    Its run record (Method) →

    41 ms51 ms1,314.4≈ 0.9
  19. KnowLine-9B-Gen2

    KnowLine-9B-Gen2 · PelaAI · decision model

    Parameters: 9B dense

    Base model: Qwen3.5-9B

    Latency: median 42 ms · p90 57 ms

    Logic 43.9% · Math 60.1%

    Its run record (Method) →

    42 ms57 ms1,217.0≈ 0.9
  20. Dexy-Decision-9B-V2

    Dexy-Decision-9B-V2 · Patrick-18 · decision model

    Parameters: 9B dense

    Base model: vllm-sr/d3-flash

    Latency: median 46 ms · p90 51 ms

    Logic 45.3% · Math 58.8%

    Its run record (Method) →

    46 ms51 ms1,272.0≈ 0.9
  21. Verdin-4B-v1.1

    Verdin-4B-v1.1 · victorbennett · decision model

    Parameters: 4B dense

    Base model: Qwen3.5-4B

    Latency: median 46 ms · p90 52 ms

    Logic 39.7% · Math 51.5%

    Its run record (Method) →

    46 ms52 ms1,240.0≈ 1.0
  22. KnowLine-12B-Gen1

    KnowLine-12B-Gen1 · PelaAI · decision model

    Parameters: 12B dense

    Base model: Gemma 4 12B

    Latency: median 51 ms · p90 76 ms

    Logic 50.6% · Math 66.3%

    Its run record (Method) →

    51 ms76 ms1,000.8≈ 0.9
  23. KnowLine-2B-Gen1

    KnowLine-2B-Gen1 · PelaAI · decision model

    Parameters: 2B dense

    Base model: Qwen3.5-2B

    Latency: median 51 ms · p90 63 ms

    Logic 31.7% · Math 35.3%

    Its run record (Method) →

    51 ms63 ms1,043.2≈ 0.9
  24. Winnow-12B

    Winnow-12B · EldanRing · decision model

    Parameters: 12B dense

    Latency: median 55 ms · p90 72 ms

    Logic 39.7% · Math 46.9%

    Its run record (Method) →

    55 ms72 ms11.5≈ 1.0
  25. d3

    d3 · vLLM Semantic Router · decision model

    Parameters: 26.09B dense

    Latency: median 57 ms · p90 86 ms

    Logic 54.5% · Math 70.4%

    Its run record (Method) →

    57 ms86 ms896.2≈ 1.0
  26. Valen-Decision-1.5-9B

    Valen-Decision-1.5-9B · Valen-Team · decision model

    Parameters: 9B dense

    Base model: Qwen3.5-9B

    Latency: median 59 ms · p90 65 ms

    Logic 44.4% · Math 55.3%

    Its run record (Method) →

    59 ms65 ms1,000.8≈ 1.0
  27. StartLux-Decision-v1.1-4B-Instant

    StartLux-Decision-v1.1-4B-Instant · StartLux · decision model

    Parameters: 4B dense

    Latency: median 68 ms · p90 80 ms

    Logic 45.9% · Math 55.8%

    Its run record (Method) →

    68 ms80 ms828.4≈ 1.0
  28. KnowLine 4B Gen5.5

    KnowLine 4B Gen5.5 · PelaAI · decision model

    Parameters: 4B dense

    Latency: median 76 ms · p90 104 ms

    Logic 42.3% · Math 56.0%

    Its run record (Method) →

    76 ms104 ms706.7≈ 1.0
  29. Elephant-1

    Elephant-1 · Base Labs · decision model

    Parameters: 9B dense

    Latency: median 80 ms · p90 94 ms

    Logic 46.4% · Math 61.3%

    Its run record (Method) →

    80 ms94 ms711.8≈ 1.0
  30. OneJev-27B

    OneJev-27B · OmniJev · decision model

    Parameters: 27B dense

    Latency: median 80 ms · p90 134 ms

    Logic 40.6% · Math 53.8%

    Its run record (Method) →

    80 ms134 ms597.5≈ 1.0
  31. GLiDE v1.2

    GLiDE v1.2 · Fastino · decision model

    Parameters: 27B dense

    Latency: median 81 ms · p90 136 ms

    Logic 55.2% · Math 69.9%

    Its run record (Method) →

    81 ms136 ms627.9≈ 1.0
  32. Drex Max

    Drex Max · Nace.AI · decision model

    Parameters: 27B dense

    Latency: median 81 ms · p90 130 ms

    Logic 53.1% · Math 70.3%

    Its run record (Method) →

    81 ms130 ms650.7≈ 1.0
  33. Perplexity Decider 27B v1.3 NVIDIA RTX PRO 6000 Blackwell

    Perplexity Decider 27B v1.3 NVIDIA RTX PRO 6000 Blackwell · Perplexity · decision model

    Parameters: 27B dense

    Latency: median 97 ms · p90 125 ms

    Logic 54.8% · Math 72.5%

    Its run record (Method) →

    97 ms125 ms583.3≈ 1.0
  34. Clef-Flash

    Clef-Flash · Cloudflare · decision model

    Parameters: 9B dense

    Latency: median 103 ms · p90 131 ms

    Logic 38.9% · Math 49.8%

    Its run record (Method) →

    103 ms131 ms552.3≈ 1.0
  35. KnowLine-31B-Gen2

    KnowLine-31B-Gen2 · PelaAI · decision model

    Parameters: 31B dense

    Base model: Gemma 4 31B

    Latency: median 116 ms · p90 177 ms

    Logic 58.9% · Math 80.3%

    Its run record (Method) →

    116 ms177 ms446.1≈ 1.0
  36. Metask-Jev-4B

    Metask-Jev-4B · Metask Lab · decision model

    Parameters: 4.54B dense

    Latency: median 117 ms · p90 146 ms

    Logic 32.1% · Math 39.2%

    Its run record (Method) →

    117 ms146 ms483.2≈ 1.0
  37. Copernicus-31B

    Copernicus-31B · Poindexter Labs · decision model

    Parameters: 31B dense

    Latency: median 128 ms · p90 218 ms

    Logic 58.7% · Math 77.3%

    Its run record (Method) →

    128 ms218 ms406.5≈ 1.0
  38. d3-ultra

    d3-ultra · vLLM Semantic Router · decision model

    Parameters: 26.09B dense

    Latency: median 129 ms · p90 168 ms

    Logic 54.6% · Math 71.5%

    Its run record (Method) →

    129 ms168 ms20.3≈ 1.0
  39. StartLux-Decision-v1.1-9B-Instant

    StartLux-Decision-v1.1-9B-Instant · StartLux · decision model

    Parameters: 9B dense

    Latency: median 142 ms · p90 165 ms

    Logic 46.6% · Math 58.4%

    Its run record (Method) →

    142 ms165 ms397.5≈ 1.0
  40. Mammoth-1

    Mammoth-1 · Base Labs · decision model

    Parameters: 27B dense

    Latency: median 150 ms · p90 232 ms

    Logic 54.5% · Math 74.0%

    Its run record (Method) →

    150 ms232 ms345.9≈ 1.0
  41. Karpo Decision 31B

    Karpo Decision 31B · MachinePulse · decision model

    Parameters: 31B dense

    Latency: median 162 ms · p90 198 ms

    Logic 60.7% · Math 81.6%

    Its run record (Method) →

    162 ms198 ms351.4≈ 1.0
  42. Cerno 9B v1.2

    Cerno 9B v1.2 · Apertix AI · decision model

    Parameters: 9B dense

    Latency: median 163 ms · p90 240 ms

    Logic 43.3% · Math 56.0%

    Its run record (Method) →

    163 ms240 ms327.5≈ 1.0
  43. Jev 1.14

    Jev 1.14 · TypeSafe · decision model

    Parameters: not disclosed by maker

    Latency: median 164 ms · p90 188 ms

    Logic 48.2% · Math 65.1%

    Its run record (Method) →

    164 ms188 ms331.4≈ 0.9
  44. Jev 1.13

    Jev 1.13 · TypeSafe · decision model

    Parameters: not disclosed by maker

    Latency: median 166 ms · p90 205 ms

    Logic 46.3% · Math 64.1%

    Its run record (Method) →

    166 ms205 ms335.3≈ 1.0
  45. d1.1

    d1.1 · Liquid AI · decision model

    Parameters: not disclosed by maker

    Latency: median 174 ms · p90 285 ms

    Logic 46.0% · Math 65.2%

    Its run record (Method) →

    174 ms285 ms286.6≈ 1.0
  46. GLiDE 9B

    GLiDE 9B · Fastino · decision model

    Parameters: 7.94B dense

    Latency: median 177 ms · p90 222 ms

    Logic 47.8% · Math 57.9%

    Its run record (Method) →

    177 ms222 ms309.5≈ 1.0
  47. d1-ultra

    d1-ultra · Liquid AI · decision model

    Parameters: not disclosed by maker (built on a larger open-weight model, not named with a count)

    Latency: median 180 ms · p90 426 ms

    Logic 54.9% · Math 74.1%

    Its run record (Method) →

    180 ms426 ms250.2≈ 1.0
  48. d1

    d1 · Liquid AI · decision model

    Parameters: not disclosed by maker

    Latency: median 185 ms · p90 292 ms

    Logic 40.8% · Math 55.6%

    Its run record (Method) →

    185 ms292 ms276.9≈ 1.0
  49. Clef

    Clef · Cloudflare · decision model

    Parameters: 27B dense

    Latency: median 185 ms · p90 216 ms

    Logic 40.8% · Math 51.1%

    Its run record (Method) →

    185 ms216 ms308.1≈ 1.0
  50. REAX 1.3

    REAX 1.3 · REAX · decision model

    Parameters: 57.36B dense

    Latency: median 200 ms · p90 260 ms

    Logic 59.3% · Math 78.0%

    Its run record (Method) →

    200 ms260 ms279.7≈ 1.0
  51. Logits Sigma-1 Duo

    Logits Sigma-1 Duo · Logits · decision model

    Parameters: 57.36B dense

    Latency: median 207 ms · p90 324 ms

    Logic 57.8% · Math 75.2%

    Its run record (Method) →

    207 ms324 ms251.5≈ 1.0
  52. d1.2

    d1.2 · Liquid AI · decision model

    Parameters: not disclosed by maker

    Latency: median 208 ms · p90 390 ms

    Logic 49.2% · Math 65.5%

    Its run record (Method) →

    208 ms390 ms232.0≈ 1.0
  53. d1.2-ultra

    d1.2-ultra · Liquid AI · decision model

    Parameters: not disclosed by maker

    Latency: median 217 ms · p90 347 ms

    Logic 51.6% · Math 69.9%

    Its run record (Method) →

    217 ms347 ms230.9≈ 1.0
  54. Gemini Decision 1 20261111

    Gemini Decision 1 20261111 · Google · decision model

    Parameters: not disclosed by maker (closed weights)

    Latency: median 227 ms · p90 347 ms

    Logic 50.0% · Math 71.3%

    Its run record (Method) →

    227 ms347 ms65.7≈ 0.4
  55. Perplexity Decider 27B v1.2 API

    Perplexity Decider 27B v1.2 API · Perplexity · decision model

    Parameters: 27B dense

    Latency: median 230 ms · p90 439 ms

    Logic 53.3% · Math 69.8%

    Its run record (Method) →

    230 ms439 ms199.4≈ 1.0
  56. Open-Jev-27B-v1.1

    Open-Jev-27B-v1.1 · Zefan Cai · decision model

    Parameters: 27B dense

    Base model: Qwen3.8-27B + adapter

    Latency: median 250 ms · p90 482 ms

    Logic 42.6% · Math 54.6%

    Its run record (Method) →

    250 ms482 ms195.6≈ 1.0
  57. GLiDE API

    GLiDE API · Fastino · decision model

    Parameters: not disclosed by maker

    Latency: median 266 ms · p90 284 ms

    Logic 54.5% · Math 70.8%

    Its run record (Method) →

    266 ms284 ms872.4≈ 4.0
  58. Mercury Decide

    Mercury Decide · Inception · decision model

    Parameters: not disclosed by maker

    Latency: median 293 ms · p90 349 ms

    Logic 45.1% · Math 53.5%

    Its run record (Method) →

    293 ms349 ms193.3≈ 1.0
  59. Turion Jev 31B

    Turion Jev 31B · Turion.AI · decision model

    Parameters: 31B dense

    Latency: median 301 ms · p90 448 ms

    Logic 52.3% · Math 69.0%

    Its run record (Method) →

    301 ms448 ms177.1≈ 1.0
  60. Microsoft-Decision-1

    Microsoft-Decision-1 · Microsoft · decision model

    Parameters: not disclosed by maker

    Base model: Qwen3.5-9B

    Latency: median 312 ms · p90 360 ms

    Logic 44.4% · Math 59.4%

    Its run record (Method) →

    312 ms360 ms184.3≈ 1.0
  61. Drex Max v1.1

    Drex Max v1.1 · Nace.AI · decision model

    Parameters: not disclosed by maker

    Latency: median 327 ms · p90 518 ms

    Logic 54.8% · Math 70.1%

    Its run record (Method) →

    327 ms518 ms118.6≈ 1.0
  62. Drex 2.0

    Drex 2.0 · Nace.AI · decision model

    Parameters: 35B MoE · 3B active

    Latency: median 339 ms · p90 472 ms

    Logic 41.9% · Math 59.0%

    Its run record (Method) →

    339 ms472 ms58.3≈ 1.0
  63. celeris-1.1-decision

    celeris-1.1-decision · Celeris · decision model

    Parameters: not disclosed by maker

    Latency: median 381 ms · p90 388 ms

    Logic 47.3% · Math 69.9%

    Its run record (Method) →

    381 ms388 ms151.9≈ 1.0
  64. SubQ Snap 1.1

    SubQ Snap 1.1 · Subquadratic · decision model

    Parameters: not disclosed by maker

    Shown once with SubQ Snap Preview: the same answer to every question

    Latency: median 382 ms · p90 391 ms

    Logic 52.4% · Math 68.6%

    Its run record (Method) →

    382 ms391 ms155.2≈ 1.0
  65. celeris-1-decision

    celeris-1-decision · Celeris · decision model

    Parameters: not disclosed by maker

    Latency: median 388 ms · p90 392 ms

    Logic 47.3% · Math 70.0%

    Its run record (Method) →

    388 ms392 ms153.4≈ 1.0
  66. Perplexity Decider 27B v1.2 API · with web

    Perplexity Decider 27B v1.2 API · with web · Perplexity · decision model

    Parameters: 27B dense

    Latency: median 430 ms · p90 707 ms

    Logic 52.5% · Math 69.3%

    Its run record (Method) →

    430 ms707 ms106.0≈ 0.9
  67. GLiDE 27B

    GLiDE 27B · Fastino · decision model

    Parameters: 27B dense

    Latency: median 450 ms · p90 587 ms

    Logic 58.4% · Math 72.0%

    Its run record (Method) →

    450 ms587 ms124.2≈ 1.0
  68. Sieve 31B v3

    Sieve 31B v3 · Sthānika AI · decision model

    Parameters: 31B dense

    Latency: median 676 ms · p90 1,173 ms

    Logic 57.9% · Math 77.1%

    Its run record (Method) →

    676 ms1,173 ms75.0≈ 1.0
  69. KnowLine-31B-Gen1.5

    KnowLine-31B-Gen1.5 · PelaAI · decision model

    Parameters: 31B dense

    Latency: median 811 ms · p90 1,111 ms

    Logic 59.0% · Math 77.4%

    Its run record (Method) →

    811 ms1,111 ms68.4≈ 1.0

Fastino Other OmniJev Liquid AI StartLux PelaAI Perplexity quincyqiang scar-ai Strands Agents Blockbrain Labs newsbang tattrongvu Patrick-18 victorbennett EldanRing vLLM Semantic Router Valen-Team Base Labs Nace.AI Cloudflare Metask Lab Poindexter Labs MachinePulse Apertix AI TypeSafe REAX Logits Google Zefan Cai Inception Turion.AI Microsoft Celeris Subquadratic Sthānika AI 90th percentile

Accuracy across latency

Each setup's accuracy against its median latency (log scale); the dotted line joins the setups no other is both faster and more accurate than; shaded: faster and more accurate than the median setup

1 ms10 ms100 ms1 s10%27.5%45%62.5%80%Median latency per question (log scale)AccuracyLower on this axis and more accurate than the median setupPareto front: no other setup is both lower on this axis and more accurate

GLiDE 0.4B: 21.5% at a median 11 ms

GLiClass v4 Edge: 23.0% at a median 12 ms

OneJev1.6-0.8B: 27.8% at a median 15 ms

d1-2B: 34.9% at a median 17 ms

GLiDE 0.8B: 31.1% at a median 20 ms

StartLux-Decision-v1.1-0.8B-Instant: 28.0% at a median 24 ms

KnowLine-0.8B-Gen1: 27.7% at a median 26 ms

GLiDE 2B: 35.2% at a median 28 ms

Perplexity Decider Flash v1.3: 52.1% at a median 29 ms

Chinese Decision Pointer 4B: 47.0% at a median 30 ms

vera-medium-omni: 52.5% at a median 31 ms

Strands Decider 2B: 26.5% at a median 35 ms

StartLux-Decision-v1.1-2B-Instant: 35.9% at a median 36 ms

Cygnet: 42.6% at a median 38 ms

Adjudex 9B v5: 56.7% at a median 38 ms

KnowLine-9B-Gen1: 54.5% at a median 38 ms

Jeffan-9B-1211: 55.8% at a median 39 ms

Perplexity Decider Flash v1.2: 51.1% at a median 41 ms

KnowLine-9B-Gen2: 54.6% at a median 42 ms

Dexy-Decision-9B-V2: 54.2% at a median 46 ms

Verdin-4B-v1.1: 47.5% at a median 46 ms

KnowLine-12B-Gen1: 60.9% at a median 51 ms

KnowLine-2B-Gen1: 34.1% at a median 51 ms

Winnow-12B: 44.4% at a median 55 ms

d3: 65.0% at a median 57 ms

Valen-Decision-1.5-9B: 51.6% at a median 59 ms

StartLux-Decision-v1.1-4B-Instant: 52.4% at a median 68 ms

KnowLine 4B Gen5.5: 51.3% at a median 76 ms

Elephant-1: 56.2% at a median 80 ms

OneJev-27B: 49.3% at a median 80 ms

GLiDE (v1.2): 64.8% at a median 81 ms

Drex Max: 64.4% at a median 81 ms

Perplexity Decider 27B v1.3 (NVIDIA RTX PRO 6000 Blackwell): 66.4% at a median 97 ms

Clef-Flash: 46.1% at a median 103 ms

KnowLine-31B-Gen2: 73.0% at a median 116 ms

Metask-Jev-4B: 36.7% at a median 117 ms

Copernicus-31B: 71.0% at a median 128 ms

d3-ultra: 65.7% at a median 129 ms

StartLux-Decision-v1.1-9B-Instant: 54.4% at a median 142 ms

Mammoth-1: 67.3% at a median 150 ms

Karpo Decision 31B: 74.4% at a median 162 ms

Cerno 9B v1.2: 51.7% at a median 163 ms

Jev 1.14: 59.3% at a median 164 ms

Jev 1.13: 58.0% at a median 166 ms

d1.1: 58.6% at a median 174 ms

GLiDE 9B: 54.5% at a median 177 ms

d1-ultra: 67.5% at a median 180 ms

d1: 50.5% at a median 185 ms

Clef: 47.6% at a median 185 ms

REAX 1.3: 71.6% at a median 200 ms

Logits Sigma-1 Duo: 69.3% at a median 207 ms

d1.2: 59.9% at a median 208 ms

d1.2-ultra: 63.6% at a median 217 ms

Gemini Decision 1 (20261111): 64.1% at a median 227 ms

Perplexity Decider 27B v1.2 (API): 64.1% at a median 230 ms

Open-Jev-27B-v1.1: 50.5% at a median 250 ms

GLiDE (API): 65.2% at a median 266 ms

Mercury Decide: 50.6% at a median 293 ms

Turion Jev 31B: 63.3% at a median 301 ms

Microsoft-Decision-1: 54.3% at a median 312 ms

Drex Max v1.1: 64.8% at a median 327 ms

Drex 2.0: 53.1% at a median 339 ms

celeris-1.1-decision: 62.1% at a median 381 ms

SubQ Snap 1.1: 63.0% at a median 382 ms

celeris-1-decision: 62.2% at a median 388 ms

Perplexity Decider 27B v1.2 (API · with web): 63.6% at a median 430 ms

GLiDE 27B: 67.3% at a median 450 ms

Sieve 31B v3: 70.5% at a median 676 ms

KnowLine-31B-Gen1.5: 71.1% at a median 811 ms

Fastino Other OmniJev Liquid AI StartLux PelaAI Perplexity quincyqiang scar-ai Strands Agents Blockbrain Labs newsbang tattrongvu Patrick-18 victorbennett EldanRing vLLM Semantic Router Valen-Team Base Labs Nace.AI Cloudflare Metask Lab Poindexter Labs MachinePulse Apertix AI TypeSafe REAX Logits Google Zefan Cai Inception Turion.AI Microsoft Celeris Subquadratic Sthānika AI

Latency is the wall time of each request from our machine; runs that shared a GPU or ran with several requests in flight slow each other down, so throughput (questions / min) depends on the run's concurrency. In flight: requests in flight on average, the sum of request latencies over the run's active time (its first to last answer with gaps of more than 10 minutes left out), e.g. “≈ 6” is six at a time.

Median latency is per request; with several requests in flight a run finishes faster than requests × median.

SubQ Snap Preview and SubQ Snap 1.1 returned the same answer to every question (and the same probabilities on 3,276 of 3,288); shown once.

Perplexity Decider 27B v1.2 (API) and Perplexity Decider 27B v1.2 (NVIDIA RTX PRO 6000 Blackwell): two setups of the same model, one on an NVIDIA RTX PRO 6000 Blackwell (96 GB) and one through Perplexity's API; not verified to be identical weights or settings. Its row is the API setup on every test; Perplexity Decider 27B v1.2 (NVIDIA RTX PRO 6000 Blackwell) is under the row's other settings and All setups.

Every setup, every measure

SetupWebAccuracy% of N (N ≤ 3,288) · Wilson 95%Tie groupHolm · McNemarRight / NquestionsKept at ≥99%right / wrongAUROCright vs wrongNon-hallucinationof answers not right, % “I don't know” · higher is betterMedian latencymsp90ms, nearest rankCost$ per 1,000 qper 1M q
Karpo Decision 31B Open · 31B

Karpo Decision 31B · MachinePulse · decision model

Parameters: 31B dense

Latency: median 162 ms · p90 198 ms

Logic 60.7% · Math 81.6%

Its run record (Method) →

—74.4%72.9%–75.9%a2,447 / 3,2881,006 right of 1,009 kept3 wrong · 2,279 skipped0.8825.7%answered 98%162 ms198 ms$0.118$118estimate · RTX PRO 6000 Blackwell
KnowLine-31B-Gen2 Open · 31B

KnowLine-31B-Gen2 · PelaAI · decision model

Parameters: 31B dense

Base model: Gemma 4 31B

Latency: median 116 ms · p90 177 ms

Logic 58.9% · Math 80.3%

Its run record (Method) →

—73.0%71.4%–74.5%ab2,399 / 3,288523 right of 524 kept1 wrong · 2,764 skipped0.8820.1%answered 100%116 ms177 ms$0.093$92.98estimate · RTX PRO 6000 Blackwell
REAX 1.3 Open · 57B

REAX 1.3 · REAX · decision model

Parameters: 57.36B dense

Latency: median 200 ms · p90 260 ms

Logic 59.3% · Math 78.0%

Its run record (Method) →

—71.6%70.0%–73.1%b2,354 / 3,288229 right of 229 kept0 mistakes · 3,059 skipped0.8783.8%answered 99%200 ms260 ms$0.148$148estimate · RTX PRO 6000 Blackwell
KnowLine-31B-Gen1.5 Open · 31B

KnowLine-31B-Gen1.5 · PelaAI · decision model

Parameters: 31B dense

Latency: median 811 ms · p90 1,111 ms

Logic 59.0% · Math 77.4%

Its run record (Method) →

—71.1%69.5%–72.6%bc2,338 / 3,288553 right of 554 kept1 wrong · 2,734 skipped0.8700.7%answered 100%811 ms1,111 ms$0.052$51.85estimate · DGX Spark
Copernicus-31B Open · 31B

Copernicus-31B · Poindexter Labs · decision model

Parameters: 31B dense

Latency: median 128 ms · p90 218 ms

Logic 58.7% · Math 77.3%

Its run record (Method) →

—71.0%69.4%–72.5%bc2,333 / 3,288384 right of 384 kept0 mistakes · 2,904 skipped0.8731.8%answered 99%128 ms218 ms$0.102$102estimate · RTX PRO 6000 Blackwell
Sieve 31B v3 Open · 31B

Sieve 31B v3 · Sthānika AI · decision model

Parameters: 31B dense

Latency: median 676 ms · p90 1,173 ms

Logic 57.9% · Math 77.1%

Its run record (Method) →

—70.5%68.9%–72.1%bcd2,319 / 3,288294 right of 296 kept2 wrong · 2,992 skipped0.8653.8%answered 99%676 ms1,173 ms$0.047$47.29estimate · DGX Spark
Logits Sigma-1 Duo Open · 57B

Logits Sigma-1 Duo · Logits · decision model

Parameters: 57.36B dense

Latency: median 207 ms · p90 324 ms

Logic 57.8% · Math 75.2%

Its run record (Method) →

—69.3%67.7%–70.8%cde2,277 / 3,28823 right of 23 kept0 mistakes · 3,265 skipped0.8684.4%answered 99%207 ms324 ms$0.165$165estimate · RTX PRO 6000 Blackwell
d1-ultra Closed · Efficient

d1-ultra · Liquid AI · decision model

Parameters: not disclosed by maker (built on a larger open-weight model, not named with a count)

Latency: median 180 ms · p90 426 ms

Logic 54.9% · Math 74.1%

Its run record (Method) →

—67.5%65.9%–69.1%defg2,221 / 3,288233 right of 233 kept0 mistakes · 3,055 skipped0.85289.1%answered 52%180 ms426 ms$0.008$8.37
GLiDE 27B Open · 27B

GLiDE 27B · Fastino · decision model

Parameters: 27B dense

Latency: median 450 ms · p90 587 ms

Logic 58.4% · Math 72.0%

Its run record (Method) →

—67.3%65.7%–68.9%efh2,214 / 3,288272 right of 274 kept2 wrong · 3,014 skipped0.85231.9%answered 89%450 ms587 ms$0.029$28.55estimate · DGX Spark
Mammoth-1 Open · 27B

Mammoth-1 · Base Labs · decision model

Parameters: 27B dense

Latency: median 150 ms · p90 232 ms

Logic 54.5% · Math 74.0%

Its run record (Method) →

—67.3%65.7%–68.9%efghi2,213 / 3,288518 right of 520 kept2 wrong · 2,768 skipped0.86871.4%answered 69%150 ms232 ms$0.028$27.98 list$0.120$120estimate · RTX PRO 6000 Blackwell
Perplexity Decider 27B v1.3 NVIDIA RTX PRO 6000 Blackwell Open · 27B

Perplexity Decider 27B v1.3 NVIDIA RTX PRO 6000 Blackwell · Perplexity · decision model

Parameters: 27B dense

Latency: median 97 ms · p90 125 ms

Logic 54.8% · Math 72.5%

Its run record (Method) →

—66.4%64.8%–68.0%fghij2,184 / 3,288385 right of 385 kept0 mistakes · 2,903 skipped0.84413.5%answered 95%97 ms125 ms$0.071$71.10estimate · RTX PRO 6000 Blackwell
d3-ultra Open · 26B

d3-ultra · vLLM Semantic Router · decision model

Parameters: 26.09B dense

Latency: median 129 ms · p90 168 ms

Logic 54.6% · Math 71.5%

Its run record (Method) →

—65.7%64.1%–67.3%fghijk2,161 / 3,28839 right of 39 kept0 mistakes · 3,249 skipped0.8357.5%answered 97%129 ms168 ms$0.096$95.71estimate · RTX PRO 6000 Blackwell
GLiDE API Closed · Efficient

GLiDE API · Fastino · decision model

Parameters: not disclosed by maker

Latency: median 266 ms · p90 284 ms

Logic 54.5% · Math 70.8%

Its run record (Method) →

—65.2%63.6%–66.8%fghijkl2,144 / 3,288213 right of 213 kept0 mistakes · 3,075 skipped0.85249.2%answered 80%266 ms284 ms$0.036$36.42
d3 Open · 26B

d3 · vLLM Semantic Router · decision model

Parameters: 26.09B dense

Latency: median 57 ms · p90 86 ms

Logic 54.5% · Math 70.4%

Its run record (Method) →

—65.0%63.3%–66.6%fghijkl2,136 / 3,288216 right of 217 kept1 wrong · 3,071 skipped0.8589.3%answered 97%57 ms86 ms$0.046$46.28estimate · RTX PRO 6000 Blackwell
Drex Max v1.1 Closed · Efficient (provisional)

Drex Max v1.1 · Nace.AI · decision model

Parameters: not disclosed by maker

Latency: median 327 ms · p90 518 ms

Logic 54.8% · Math 70.1%

Its run record (Method) →

—64.8%63.2%–66.5%fghijkl2,132 / 3,288404 right of 405 kept1 wrong · 2,883 skipped0.84648.3%answered 79%327 ms518 ms$0.016$16.13
GLiDE v1.2 Open · 27B

GLiDE v1.2 · Fastino · decision model

Parameters: 27B dense

Latency: median 81 ms · p90 136 ms

Logic 55.2% · Math 69.9%

Its run record (Method) →

—64.8%63.2%–66.5%gijkl2,132 / 3,288176 right of 176 kept0 mistakes · 3,112 skipped0.84342.3%answered 83%81 ms136 ms$0.010$9.71 list$0.066$66.05estimate · RTX PRO 6000 Blackwell
Drex Max Open · 27B

Drex Max · Nace.AI · decision model

Parameters: 27B dense

Latency: median 81 ms · p90 130 ms

Logic 53.1% · Math 70.3%

Its run record (Method) →

—64.4%62.8%–66.0%gijkl2,118 / 3,288866 right of 884 kept18 wrong · 2,404 skipped0.8415.7%answered 98%81 ms130 ms$0.016$16.31 list$0.064$63.74estimate · RTX PRO 6000 Blackwell
Perplexity Decider 27B v1.2 API Open · 27B

Perplexity Decider 27B v1.2 API · Perplexity · decision model

Parameters: 27B dense

Latency: median 230 ms · p90 439 ms

Logic 53.3% · Math 69.8%

Its run record (Method) →

—64.1%62.5%–65.8%jklm2,109 / 3,288409 right of 411 kept2 wrong · 2,877 skipped0.84910.3%answered 96%230 ms439 ms$0.005$5.38
Gemini Decision 1 20261111 Closed · Efficient

Gemini Decision 1 20261111 · Google · decision model

Parameters: not disclosed by maker (closed weights)

Latency: median 227 ms · p90 347 ms

Logic 50.0% · Math 71.3%

Its run record (Method) →

—64.1%62.4%–65.7%hijklm2,106 / 3,288257 right of 257 kept0 mistakes · 3,031 skipped0.8383.0%answered 99%227 ms347 ms$0.027$27.07
d1.2-ultra Closed · Undisclosed

d1.2-ultra · Liquid AI · decision model

Parameters: not disclosed by maker

Latency: median 217 ms · p90 347 ms

Logic 51.6% · Math 69.9%

Its run record (Method) →

—63.6%61.9%–65.2%jklm2,091 / 3,288128 right of 128 kept0 mistakes · 3,160 skipped0.85690.2%answered 48%217 ms347 ms$0.008$8.37
Perplexity Decider 27B v1.2 API · with web Open · 27B

Perplexity Decider 27B v1.2 API · with web · Perplexity · decision model

Parameters: 27B dense

Latency: median 430 ms · p90 707 ms

Logic 52.5% · Math 69.3%

Its run record (Method) →

with web63.6%61.9%–65.2%klm2,090 / 3,288404 right of 405 kept1 wrong · 2,883 skipped0.85425.8%answered 90%430 ms707 ms$1.025$1,025
Turion Jev 31B Open · 31B

Turion Jev 31B · Turion.AI · decision model

Parameters: 31B dense

Latency: median 301 ms · p90 448 ms

Logic 52.3% · Math 69.0%

Its run record (Method) →

—63.3%61.6%–64.9%jklmn2,081 / 3,28850 right of 50 kept0 mistakes · 3,238 skipped0.8441.9%answered 99%301 ms448 ms$0.020$20.01estimate · DGX Spark
SubQ Snap 1.1 Closed · Efficient

SubQ Snap 1.1 · Subquadratic · decision model

Parameters: not disclosed by maker

Shown once with SubQ Snap Preview: the same answer to every question

Latency: median 382 ms · p90 391 ms

Logic 52.4% · Math 68.6%

Its run record (Method) →

—63.0%61.4%–64.7%klmn2,072 / 3,288430 right of 433 kept3 wrong · 2,855 skipped0.82758.3%answered 73%382 ms391 ms$0.011$10.75
celeris-1-decision Closed · Efficient

celeris-1-decision · Celeris · decision model

Parameters: not disclosed by maker

Latency: median 388 ms · p90 392 ms

Logic 47.3% · Math 70.0%

Its run record (Method) →

—62.2%60.6%–63.9%lmno2,046 / 3,288No answers at this threshold0.82987.2%answered 50%388 ms392 ms$0.011$11.25
celeris-1.1-decision Closed · Efficient

celeris-1.1-decision · Celeris · decision model

Parameters: not disclosed by maker

Latency: median 381 ms · p90 388 ms

Logic 47.3% · Math 69.9%

Its run record (Method) →

—62.1%60.5%–63.8%lmnop2,043 / 3,288No answers at this threshold0.83086.7%answered 51%381 ms388 ms$0.001$1.13
KnowLine-12B-Gen1 Open · 12B

KnowLine-12B-Gen1 · PelaAI · decision model

Parameters: 12B dense

Base model: Gemma 4 12B

Latency: median 51 ms · p90 76 ms

Logic 50.6% · Math 66.3%

Its run record (Method) →

—60.9%59.2%–62.5%mnopq2,002 / 3,28813 right of 13 kept0 mistakes · 3,275 skipped0.8104.5%answered 98%51 ms76 ms$0.041$41.44estimate · RTX PRO 6000 Blackwell
d1.2 Closed · Undisclosed

d1.2 · Liquid AI · decision model

Parameters: not disclosed by maker

Latency: median 208 ms · p90 390 ms

Logic 49.2% · Math 65.5%

Its run record (Method) →

—59.9%58.2%–61.6%nopqr1,970 / 3,288169 right of 169 kept0 mistakes · 3,119 skipped0.81266.9%answered 65%208 ms390 ms$0.004$3.98
Jev 1.14 Closed · Efficient

Jev 1.14 · TypeSafe · decision model

Parameters: not disclosed by maker

Latency: median 164 ms · p90 188 ms

Logic 48.2% · Math 65.1%

Its run record (Method) →

—59.3%57.6%–61.0%opqrs1,951 / 3,288263 right of 263 kept0 mistakes · 3,025 skipped0.81015.4%answered 93%164 ms188 ms$0.021$21.27
d1.1 Closed · Efficient

d1.1 · Liquid AI · decision model

Parameters: not disclosed by maker

Latency: median 174 ms · p90 285 ms

Logic 46.0% · Math 65.2%

Its run record (Method) →

—58.6%56.9%–60.3%pqrs1,928 / 3,288231 right of 231 kept0 mistakes · 3,057 skipped0.81186.0%answered 46%174 ms285 ms$0.004$3.98
Jev 1.13 Closed · Efficient

Jev 1.13 · TypeSafe · decision model

Parameters: not disclosed by maker

Latency: median 166 ms · p90 205 ms

Logic 46.3% · Math 64.1%

Its run record (Method) →

—58.0%56.3%–59.7%qrst1,907 / 3,288313 right of 313 kept0 mistakes · 2,975 skipped0.81710.2%answered 96%166 ms205 ms$0.021$21.27
Adjudex 9B v5 Open · 9.0B

Adjudex 9B v5 · newsbang · decision model

Parameters: 9B dense

Base model: Qwen3.5-9B

Latency: median 38 ms · p90 45 ms

Logic 46.4% · Math 62.1%

Its run record (Method) →

—56.7%55.0%–58.4%rstu1,865 / 3,288251 right of 253 kept2 wrong · 3,035 skipped0.8027.0%answered 97%38 ms45 ms$0.028$28.19estimate · RTX PRO 6000 Blackwell
Elephant-1 Open · 9.0B

Elephant-1 · Base Labs · decision model

Parameters: 9B dense

Latency: median 80 ms · p90 94 ms

Logic 46.4% · Math 61.3%

Its run record (Method) →

—56.2%54.5%–57.9%stu1,849 / 3,288235 right of 236 kept1 wrong · 3,052 skipped0.79769.4%answered 60%80 ms94 ms$0.010$9.79 list$0.058$58.27estimate · RTX PRO 6000 Blackwell
Jeffan-9B-1211 Open · 9.0B

Jeffan-9B-1211 · tattrongvu · decision model

Parameters: 9B dense

Base model: Qwen3.5-9B

Latency: median 39 ms · p90 53 ms

Logic 48.2% · Math 59.8%

Its run record (Method) →

—55.8%54.1%–57.5%stu1,836 / 3,28874 right of 74 kept0 mistakes · 3,214 skipped0.79813.5%answered 94%39 ms53 ms$0.032$32.40estimate · RTX PRO 6000 Blackwell
KnowLine-9B-Gen2 Open · 9.0B

KnowLine-9B-Gen2 · PelaAI · decision model

Parameters: 9B dense

Base model: Qwen3.5-9B

Latency: median 42 ms · p90 57 ms

Logic 43.9% · Math 60.1%

Its run record (Method) →

—54.6%52.9%–56.3%tuv1,794 / 3,288No answers at this threshold0.78610.4%answered 95%42 ms57 ms$0.034$34.08estimate · RTX PRO 6000 Blackwell
GLiDE 9B Open · 7.9B

GLiDE 9B · Fastino · decision model

Parameters: 7.94B dense

Latency: median 177 ms · p90 222 ms

Logic 47.8% · Math 57.9%

Its run record (Method) →

—54.5%52.8%–56.2%uv1,791 / 3,288105 right of 105 kept0 mistakes · 3,183 skipped0.78626.9%answered 86%177 ms222 ms$0.011$11.45estimate · DGX Spark
KnowLine-9B-Gen1 Open · 9.0B

KnowLine-9B-Gen1 · PelaAI · decision model

Parameters: 9B dense

Base model: Qwen3.5-9B

Latency: median 38 ms · p90 55 ms

Logic 44.3% · Math 59.8%

Its run record (Method) →

—54.5%52.8%–56.2%uv1,791 / 3,288153 right of 153 kept0 mistakes · 3,135 skipped0.7823.4%answered 98%38 ms55 ms$0.030$29.87estimate · RTX PRO 6000 Blackwell
StartLux-Decision-v1.1-9B-Instant Open · 9.0B

StartLux-Decision-v1.1-9B-Instant · StartLux · decision model

Parameters: 9B dense

Latency: median 142 ms · p90 165 ms

Logic 46.6% · Math 58.4%

Its run record (Method) →

—54.4%52.7%–56.1%uv1,788 / 3,28863 right of 63 kept0 mistakes · 3,225 skipped0.7799.6%answered 95%142 ms165 ms$0.009$8.92estimate · DGX Spark
Microsoft-Decision-1 Closed · Efficient

Microsoft-Decision-1 · Microsoft · decision model

Parameters: not disclosed by maker

Base model: Qwen3.5-9B

Latency: median 312 ms · p90 360 ms

Logic 44.4% · Math 59.4%

Its run record (Method) →

—54.3%52.6%–56.0%uvw1,784 / 3,28847 right of 47 kept0 mistakes · 3,241 skipped0.76911.4%answered 94%312 ms360 ms$0.009$9.07
Dexy-Decision-9B-V2 Open · 9.0B

Dexy-Decision-9B-V2 · Patrick-18 · decision model

Parameters: 9B dense

Base model: vllm-sr/d3-flash

Latency: median 46 ms · p90 51 ms

Logic 45.3% · Math 58.8%

Its run record (Method) →

—54.2%52.5%–55.9%uvw1,782 / 3,288263 right of 266 kept3 wrong · 3,022 skipped0.7956.1%answered 97%46 ms51 ms$0.033$32.61estimate · RTX PRO 6000 Blackwell
Drex 2.0 Open · 35B MoE · 3.0B active

Drex 2.0 · Nace.AI · decision model

Parameters: 35B MoE · 3B active

Latency: median 339 ms · p90 472 ms

Logic 41.9% · Math 59.0%

Its run record (Method) →

—53.1%51.4%–54.8%uvwx1,747 / 3,288207 right of 207 kept0 mistakes · 3,081 skipped0.78928.1%answered 85%339 ms472 ms$0.012$12.17
vera-medium-omni Open · 4.2B

vera-medium-omni · scar-ai · decision model

Parameters: 4.2B dense

Base model: Qwen3.5-4B

Latency: median 31 ms · p90 36 ms

Logic 44.1% · Math 56.8%

Its run record (Method) →

—52.5%50.8%–54.2%vwxy1,725 / 3,28839 right of 39 kept0 mistakes · 3,249 skipped0.7777.9%answered 96%31 ms36 ms$0.024$24.19estimate · RTX PRO 6000 Blackwell
StartLux-Decision-v1.1-4B-Instant Open · 4.0B

StartLux-Decision-v1.1-4B-Instant · StartLux · decision model

Parameters: 4B dense

Latency: median 68 ms · p90 80 ms

Logic 45.9% · Math 55.8%

Its run record (Method) →

—52.4%50.7%–54.1%vwxy1,724 / 3,288117 right of 117 kept0 mistakes · 3,171 skipped0.76314.4%answered 93%68 ms80 ms$0.004$4.28estimate · DGX Spark
Perplexity Decider Flash v1.3 Open · 4.0B

Perplexity Decider Flash v1.3 · Perplexity · decision model

Parameters: 4B dense

Latency: median 29 ms · p90 35 ms

Logic 44.7% · Math 55.9%

Its run record (Method) →

—52.1%50.4%–53.8%vwxy1,713 / 3,288130 right of 131 kept1 wrong · 3,157 skipped0.78215.0%answered 92%29 ms35 ms$0.024$23.77estimate · RTX PRO 6000 Blackwell
Cerno 9B v1.2 Open · 9.0B

Cerno 9B v1.2 · Apertix AI · decision model

Parameters: 9B dense

Latency: median 163 ms · p90 240 ms

Logic 43.3% · Math 56.0%

Its run record (Method) →

—51.7%50.0%–53.4%vwxy1,699 / 3,288270 right of 281 kept11 wrong · 3,007 skipped0.7763.1%answered 99%163 ms240 ms$0.011$10.82estimate · DGX Spark
Valen-Decision-1.5-9B Open · 9.0B

Valen-Decision-1.5-9B · Valen-Team · decision model

Parameters: 9B dense

Base model: Qwen3.5-9B

Latency: median 59 ms · p90 65 ms

Logic 44.4% · Math 55.3%

Its run record (Method) →

—51.6%49.9%–53.3%vwxy1,696 / 3,288165 right of 167 kept2 wrong · 3,121 skipped0.7657.1%answered 96%59 ms65 ms$0.041$41.44estimate · RTX PRO 6000 Blackwell
KnowLine 4B Gen5.5 Open · 4.0B

KnowLine 4B Gen5.5 · PelaAI · decision model

Parameters: 4B dense

Latency: median 76 ms · p90 104 ms

Logic 42.3% · Math 56.0%

Its run record (Method) →

—51.3%49.6%–53.0%vwxyz1,687 / 3,28871 right of 72 kept1 wrong · 3,216 skipped0.7524.4%answered 98%76 ms104 ms$0.005$5.02estimate · DGX Spark
Perplexity Decider Flash v1.2 Open · 4.0B

Perplexity Decider Flash v1.2 · Perplexity · decision model

Parameters: 4B dense

Latency: median 41 ms · p90 51 ms

Logic 43.1% · Math 55.2%

Its run record (Method) →

—51.1%49.4%–52.8%vwxyz1,680 / 3,288185 right of 190 kept5 wrong · 3,098 skipped0.76918.9%answered 90%41 ms51 ms$0.003$2.72 list$0.032$31.55estimate · RTX PRO 6000 Blackwell
Mercury Decide Closed · Efficient

Mercury Decide · Inception · decision model

Parameters: not disclosed by maker

Latency: median 293 ms · p90 349 ms

Logic 45.1% · Math 53.5%

Its run record (Method) →

—50.6%48.9%–52.3%wxyzA1,664 / 3,288356 right of 372 kept16 wrong · 2,916 skipped0.77710.1%answered 95%293 ms349 ms$0.005$5.46
d1 Closed · Efficient

d1 · Liquid AI · decision model

Parameters: not disclosed by maker

Latency: median 185 ms · p90 292 ms

Logic 40.8% · Math 55.6%

Its run record (Method) →

—50.5%48.8%–52.3%wxyzA1,662 / 3,288207 right of 208 kept1 wrong · 3,080 skipped0.7576.4%answered 97%185 ms292 ms$0.004$3.98
Open-Jev-27B-v1.1 Open · 27B

Open-Jev-27B-v1.1 · Zefan Cai · decision model

Parameters: 27B dense

Base model: Qwen3.8-27B + adapter

Latency: median 250 ms · p90 482 ms

Logic 42.6% · Math 54.6%

Its run record (Method) →

—50.5%48.7%–52.2%xyzA1,659 / 3,288No answers at this threshold0.75914.5%answered 92%250 ms482 ms$0.212$212estimate · RTX PRO 6000 Blackwell
OneJev-27B Open · 27B

OneJev-27B · OmniJev · decision model

Parameters: 27B dense

Latency: median 80 ms · p90 134 ms

Logic 40.6% · Math 53.8%

Its run record (Method) →

—49.3%47.6%–51.0%yzAB1,620 / 3,28830 right of 30 kept0 mistakes · 3,258 skipped0.7701.6%answered 99%80 ms134 ms$0.069$69.42estimate · RTX PRO 6000 Blackwell
Clef Open · 27B

Clef · Cloudflare · decision model

Parameters: 27B dense

Latency: median 185 ms · p90 216 ms

Logic 40.8% · Math 51.1%

Its run record (Method) →

—47.6%45.9%–49.3%zABC1,564 / 3,28818 right of 18 kept0 mistakes · 3,270 skipped0.7472.4%answered 99%185 ms216 ms$0.081$81.46 list$0.135$135estimate · RTX PRO 6000 Blackwell
Verdin-4B-v1.1 Open · 4.0B

Verdin-4B-v1.1 · victorbennett · decision model

Parameters: 4B dense

Base model: Qwen3.5-4B

Latency: median 46 ms · p90 52 ms

Logic 39.7% · Math 51.5%

Its run record (Method) →

—47.5%45.8%–49.2%ABC1,561 / 3,288161 right of 165 kept4 wrong · 3,123 skipped0.7478.9%answered 95%46 ms52 ms$0.033$33.45estimate · RTX PRO 6000 Blackwell
Chinese Decision Pointer 4B Open · 4.0B

Chinese Decision Pointer 4B · quincyqiang · decision model

Parameters: 4B dense

Base model: Qwen3.5-4B

Latency: median 30 ms · p90 42 ms

Logic 40.4% · Math 50.4%

Its run record (Method) →

—47.0%45.3%–48.7%ABC1,545 / 3,288188 right of 194 kept6 wrong · 3,094 skipped0.7437.8%answered 96%30 ms42 ms$0.025$25.03estimate · RTX PRO 6000 Blackwell
Clef-Flash Open · 9.0B

Clef-Flash · Cloudflare · decision model

Parameters: 9B dense

Latency: median 103 ms · p90 131 ms

Logic 38.9% · Math 49.8%

Its run record (Method) →

—46.1%44.4%–47.8%BCD1,515 / 3,288No answers at this threshold0.7076.9%answered 96%103 ms131 ms$0.013$12.90 list$0.075$75.10estimate · RTX PRO 6000 Blackwell
Winnow-12B Open · 12B

Winnow-12B · EldanRing · decision model

Parameters: 12B dense

Latency: median 55 ms · p90 72 ms

Logic 39.7% · Math 46.9%

Its run record (Method) →

—44.4%42.7%–46.1%CD1,461 / 3,28852 right of 53 kept1 wrong · 3,235 skipped0.73717.1%answered 90%55 ms72 ms$0.043$42.91estimate · RTX PRO 6000 Blackwell
Cygnet Open · 12B

Cygnet · Blockbrain Labs · decision model

Parameters: 11.95B dense

Base model: Gemma 4 12B IT

Latency: median 38 ms · p90 44 ms

Logic 39.3% · Math 44.3%

Its run record (Method) →

—42.6%40.9%–44.3%D1,401 / 3,288No answers at this threshold0.7297.3%answered 96%38 ms44 ms$0.029$29.34 list$0.029$28.61estimate · RTX PRO 6000 Blackwell
Metask-Jev-4B Open · 4.5B

Metask-Jev-4B · Metask Lab · decision model

Parameters: 4.54B dense

Latency: median 117 ms · p90 146 ms

Logic 32.1% · Math 39.2%

Its run record (Method) →

—36.7%35.1%–38.4%E1,208 / 3,288No answers at this threshold0.6592.0%answered 99%117 ms146 ms$0.086$85.83estimate · RTX PRO 6000 Blackwell
StartLux-Decision-v1.1-2B-Instant Open · 2.0B

StartLux-Decision-v1.1-2B-Instant · StartLux · decision model

Parameters: 2B dense

Latency: median 36 ms · p90 42 ms

Logic 30.8% · Math 38.7%

Its run record (Method) →

—35.9%34.3%–37.6%E1,182 / 3,2884 right of 4 kept0 mistakes · 3,284 skipped0.67516.4%answered 89%36 ms42 ms$0.002$2.36estimate · DGX Spark
GLiDE 2B Open · 1.9B

GLiDE 2B · Fastino · decision model

Parameters: 1.88B dense

Latency: median 28 ms · p90 37 ms

Logic 32.3% · Math 36.8%

Its run record (Method) →

—35.2%33.6%–36.9%E1,158 / 3,28829 right of 30 kept1 wrong · 3,258 skipped0.63723.1%answered 84%28 ms37 ms$0.002$1.94estimate · DGX Spark
d1-2B Open · 1.9B

d1-2B · Liquid AI · decision model

Parameters: 1.92B dense

Base model: LFM3-VL-2B

Latency: median 17 ms · p90 21 ms

Logic 30.7% · Math 37.1%

Its run record (Method) →

—34.9%33.3%–36.5%EF1,147 / 3,2887 right of 7 kept0 mistakes · 3,281 skipped0.64882.5%answered 30%17 ms21 ms$0.014$14.30estimate · RTX PRO 6000 Blackwell
KnowLine-2B-Gen1 Open · 2.0B

KnowLine-2B-Gen1 · PelaAI · decision model

Parameters: 2B dense

Base model: Qwen3.5-2B

Latency: median 51 ms · p90 63 ms

Logic 31.7% · Math 35.3%

Its run record (Method) →

—34.1%32.5%–35.7%EF1,120 / 3,28816 right of 16 kept0 mistakes · 3,272 skipped0.6266.8%answered 95%51 ms63 ms$0.003$3.40estimate · DGX Spark
GLiDE 0.8B Open · 0.8B

GLiDE 0.8B · Fastino · decision model

Parameters: 0.75B dense

Latency: median 20 ms · p90 25 ms

Logic 29.2% · Math 32.1%

Its run record (Method) →

—31.1%29.5%–32.7%FG1,022 / 3,28813 right of 13 kept0 mistakes · 3,275 skipped0.60222.7%answered 83%20 ms25 ms$0.001$1.38estimate · DGX Spark
StartLux-Decision-v1.1-0.8B-Instant Open · 0.8B

StartLux-Decision-v1.1-0.8B-Instant · StartLux · decision model

Parameters: 0.8B dense

Base model: Qwen3.5-0.8B

Latency: median 24 ms · p90 29 ms

Logic 24.7% · Math 29.7%

Its run record (Method) →

—28.0%26.5%–29.5%GH920 / 3,288No answers at this threshold0.60514.9%answered 89%24 ms29 ms$0.002$1.62estimate · DGX Spark
OneJev1.6-0.8B Open · 0.8B

OneJev1.6-0.8B · OmniJev · decision model

Parameters: 0.8B dense

Latency: median 15 ms · p90 17 ms

Logic 27.0% · Math 28.2%

Its run record (Method) →

—27.8%26.3%–29.4%GH915 / 3,2882 right of 2 kept0 mistakes · 3,286 skipped0.5762.0%answered 99%15 ms17 ms$0.001$1.11estimate · DGX Spark
KnowLine-0.8B-Gen1 Open · 0.8B

KnowLine-0.8B-Gen1 · PelaAI · decision model

Parameters: 0.8B dense

Base model: Qwen3.5-0.8B

Latency: median 26 ms · p90 35 ms

Logic 26.7% · Math 28.2%

Its run record (Method) →

—27.7%26.2%–29.2%GH910 / 3,2882 right of 2 kept0 mistakes · 3,286 skipped0.5900.6%answered 100%26 ms35 ms$0.002$1.80estimate · DGX Spark
Strands Decider 2B Open · 2.0B

Strands Decider 2B · Strands Agents · decision model

Parameters: 2B dense

Latency: median 35 ms · p90 42 ms

Logic 22.4% · Math 28.7%

Its run record (Method) →

—26.5%25.0%–28.1%HI872 / 3,288No answers at this threshold0.5735.0%answered 96%35 ms42 ms$0.027$26.72estimate · RTX PRO 6000 Blackwell
GLiClass v4 Edge Open · 0.4B

GLiClass v4 Edge · decision model

Parameters: 0.4B dense

Base model: DeBERTa-v3-large

Latency: median 12 ms · p90 23 ms

Logic 23.6% · Math 22.7%

Its run record (Method) →

—23.0%21.6%–24.4%IJ755 / 3,288No answers at this threshold0.51115.6%answered 87%12 ms23 ms$0.001$1.01estimate · DGX Spark
GLiDE 0.4B Open · 0.4B

GLiDE 0.4B · Fastino · decision model

Parameters: 0.40B dense

Latency: median 11 ms · p90 14 ms

Logic 23.2% · Math 20.6%

Its run record (Method) →

—21.5%20.1%–22.9%J706 / 3,288No answers at this threshold0.52073.9%answered 32%11 ms14 ms$0.001$0.83estimate · DGX Spark

Tie groups and “not measured”

estimate · RTX PRO 6000 Blackwell / DGX Spark = run on our own hardware; its cost is an estimate, explained on hover, focus or tap of the amount: the machine's wall-clock time on the run, first to last answer while it was running (pauses, estimated from gaps of more than 10 minutes between answers, left out and listed), × $2.49/h per NVIDIA RTX PRO 6000 Blackwell (96 GB) (Runpod's price for the card) or × $0.213/h per NVIDIA DGX Spark (GB10) (its purchase over three years plus electricity; see Method), each run charged in full even when runs shared the machine, so at most; sorting and the cost chart use the estimate; list = maker's list price × our token counts; unmarked = provider's price × usage the API reported.