Benchmark Almanac

TaxEval v2

Muse Spark 1.2 80.4% leads since Aug 2026 Not in model cards

Models answer expert-written tax questions. A judge grades answer correctness and stepwise reasoning against expert solutions.

Over 1,500 questions: 20 public on request, 300 licensable validation and 1,223 private test. Vals archived it as saturated by 2026-09-01. Release month not stated on the page.

Made by Vals AI Finance Tasks kept private Last model added Sep 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Open-ended Free-form text, documents, designs or media judged as a whole.
Leaderboard
131
Entries on its own leaderboard, 129 models we track

Muse Spark 1.2 leads with 80.4% among models released by Sep 2026

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

The newest model with a score here came out in Sep 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

131 entries. Scroll sideways for the rest.

  1. 80.4 Muse Spark 1.2
  2. 79.7 Muse Spark 1.1
  3. 77.7 Muse Spark
  4. 77.1 Claude Sonnet 4.6
  5. 76.9 Claude Fable 5
  6. 76.2 GPT-5.6 Luna
  7. 76.2 GPT-5.6 Terra
  8. 76.0 Claude Opus 4.6
  9. 76.0 Claude Fable 5.1
  10. 75.9 Grok 3
  11. 75.8 GPT-5.2
  12. 75.7 Kimi K3
  13. 75.7 Grok 4 Fast
  14. 75.6 Claude Opus 4.8
  15. 75.6 Claude Sonnet 5
  16. 75.6 GLM-5.3-Flash
  17. 75.5 Qwen 3.8 Max
  18. 75.5 Inkling-Small
  19. 75.3 Qwen3.7-Max
  20. 75.3 Inkling
  21. 75.3 Claude Opus 4.7
  22. 75.2 GPT-5 mini
  23. 75.1 Claude Opus 5
  24. 75.1 GPT-4.1
  25. 75.0 GPT-5.5
  26. 74.9 GPT-5.1
  27. 74.9 Claude Opus 4.5
  28. 74.9 Gemini 3.6 Flash
  29. 74.8 o4-mini
  30. 74.8 GPT-5.6 Sol
  31. 74.7 Qwen 3.6 Plus
  32. 74.7 Gemini 3.7 Flash
  33. 74.7 Kimi K2.6
  34. 74.6 o3
  35. 74.5 GPT-4o
  36. 74.5 Gemini 3.8 Flash
  37. 74.4 Gemini 3.5 Flash
  38. 74.3 o1
  39. 74.2 Kimi K2.5
  40. 74.1 Grok 4.20
  41. 74.0 Claude 3.7 Sonnet
  42. 74.0 Qwen3-Max
  43. 74.0 GPT-5.4
  44. 73.9 Gemini 3 Flash
  45. 73.8 MiMo-V2.5-Pro
  46. 73.7 Claude Opus 4.1
  47. 73.4 GPT-5
  48. 73.3 GLM-5.2
  49. 73.3 Claude Sonnet 4.5
  50. 73.1 Grok 4.1 Fast
  51. 73.1 Nemotron 3 Ultra
  52. 73.1 Mistral Large 3
  53. 73.1 DeepSeek V4 Pro 0813
  54. 73.0 Grok-3 mini
  55. 72.9 Gemini 2.5 Pro
  56. 72.9 Gemini 3.1 Pro
  57. 72.7 Gemini 2.5 Flash
  58. 72.7 MiniMax-M3
  59. 72.6 Gemini 3.5 Flash-Lite
  60. 72.6 Gemini 3 Pro
  61. 72.4 GLM-4.5
  62. 72.4 GLM-5.3
  63. 72.3 DeepSeek-R1
  64. 72.2 Qwen 3.5 Flash (hosted 35B-A3B)
  65. 72.1 DeepSeek-V4-Pro
  66. 72.0 Claude Sonnet 4
  67. 71.9 GPT-4.1 mini
  68. 71.9 Claude Opus 4
  69. 71.8 MiMo-V2.5
  70. 71.8 Gemini 3.1 Flash-Lite
  71. 71.7 Kimi K2 Thinking
  72. 71.7 Grok 4.5
  73. 71.6 gpt-oss-120b
  74. 71.3 Qwen3.6 27B
  75. 71.2 GPT-5.4 Mini
  76. 71.2 GLM-5.1
  77. 71.2 Gemini 2.5 Flash
  78. 71.1 DeepSeek-V3
  79. 71.1 Grok 4.6
  80. 70.8 Qwen 3.8 27B
  81. 70.8 Grok 4.3
  82. 70.7 DeepSeek V4 Flash 0731
  83. 70.7 Qwen3-235B-A22B
  84. 70.7 Ling-3.0-flash
  85. 70.3 Mistral Medium 3.1
  86. 70.2 Kimi K2
  87. 70.2 Claude 3.5 Sonnet
  88. 70.0 GLM-5
  89. 69.8 Gemini 2.0 Flash Thinking
  90. 69.4 o3-mini
  91. 68.8 GLM-4.7
  92. 68.2 Gemini 2.0 Pro
  93. 68.2 DeepSeek-V3.2
  94. 68.2 MiniMax-M2.5
  95. 68.0 Mistral Medium 3.5
  96. 67.7 Gemini 2.0 Flash
  97. 67.5 Claude Haiku 4.5
  98. 67.4 GPT-5.4 Nano
  99. 67.4 GPT-5 nano
  100. 67.0 Grok-2
  101. 66.6 Llama 4 Maverick
  102. 66.6 MiniMax-M2.7
  103. 66.3 MiniMax-M2.1
  104. 66.2 Gemini 2.5 Flash-Lite
  105. 66.2 GLM-4.6
  106. 65.1 Grok 4
  107. 63.8 Mistral Large 2
  108. 63.7 Llama 3.3 Nemotron Super (Thinking)
  109. 63.7 gpt-oss-20b
  110. 61.9 Magistral Medium 1.2
  111. 61.4 Cohere Command A
  112. 60.9 Llama 3.1-405B
  113. 60.9 Jamba 1.6 Large
  114. 60.8 GPT-4.1 nano
  115. 60.5 GPT-4o mini
  116. 60.3 Magistral Small 1.2
  117. 60.2 Llama 3.3 Nemotron Super (non-thinking)
  118. 59.5 Gemini 1.5 Pro
  119. 59.4 Llama 3.3 70B
  120. 59.0 Laguna XS.2
  121. 58.3 Mistral Small 3.1
  122. 58.2 Jamba 1.5 Large
  123. 57.4 Claude 3.5 Haiku
  124. 56.2 Llama 3.1-70B
  125. 55.2 Llama 4 Scout
  126. 49.1 mistral-small-2402
  127. 48.2 Gemini 1.5 Flash
  128. 44.6 Jamba 1.6 Mini
  129. 41.9 Jamba 1.5 Mini
  130. 32.3 Llama 3.1-8B
  131. 1.64 Laguna M.1
  1. 1 Muse Spark 1.2 80.4%
  2. 2 Muse Spark 1.1 79.7%
  3. 3 Muse Spark 77.7%
  4. 4 Claude Sonnet 4.6 77.1%
  5. 5 Claude Fable 5 76.9%
  6. 6 GPT-5.6 Luna 76.2%
  7. 7 GPT-5.6 Terra 76.2%
  8. 8 Claude Opus 4.6 thinking 76%
  9. 9 Claude Fable 5.1 76%
  10. 10 Grok 3 75.9%
  11. 11 GPT-5.2 75.8%
  12. 12 Kimi K3 75.7%
  13. 13 Grok 4 Fast reasoning 75.7%
  14. 14 Claude Opus 4.8 75.6%
  15. 15 Claude Sonnet 5 75.6%
  16. 16 GLM-5.3-Flash 75.6%
  17. 17 Qwen 3.8 Max 75.5%
  18. 18 Inkling-Small 75.5%
  19. 19 Qwen3.7-Max 75.3%
  20. 20 Inkling 75.3%
  21. 21 Claude Opus 4.7 75.3%
  22. 22 GPT-5 mini 75.2%
  23. 23 Claude Opus 5 75.1%
  24. 24 GPT-4.1 75.1%
  25. 25 GPT-5.5 75%
  26. 26 GPT-5.1 74.9%
  27. 27 Claude Opus 4.5 thinking 74.9%
  28. 28 Gemini 3.6 Flash 74.9%
  29. 29 o4-mini 74.8%
  30. 30 GPT-5.6 Sol 74.8%
  31. 31 Qwen 3.6 Plus 74.7%
  32. 32 Gemini 3.7 Flash 74.7%
  33. 33 Kimi K2.6 74.7%
  34. 34 o3 74.6%
  35. 35 GPT-4o Aug 2024 74.5%
  36. 36 Gemini 3.8 Flash 74.5%
  37. 37 Gemini 3.5 Flash 74.4%
  38. 38 o1 74.3%
  39. 39 Kimi K2.5 thinking 74.2%
  40. 40 Grok 4.20 reasoning 74.1%
  41. 41 Claude 3.7 Sonnet thinking 74%
  42. 42 Qwen3-Max 74%
  43. 43 GPT-5.4 xhigh 74%
  44. 44 Gemini 3 Flash 73.9%
  45. 45 MiMo-V2.5-Pro 73.8%
  46. 46 Claude Opus 4.1 thinking 73.7%
  47. 47 GPT-5 73.4%
  48. 48 GLM-5.2 73.3%
  49. 49 Claude Sonnet 4.5 thinking 73.3%
  50. 50 Grok 4.1 Fast reasoning 73.1%
  51. 51 Nemotron 3 Ultra 73.1%
  52. 52 Mistral Large 3 73.1%
  53. 53 DeepSeek V4 Pro 0813 73.1%
  54. 54 Grok-3 mini reasoning 73%
  55. 55 Gemini 2.5 Pro Jun 2025 72.9%
  56. 56 Gemini 3.1 Pro 72.9%
  57. 57 Gemini 2.5 Flash Sep 2025, non-thinking 72.7%
  58. 58 MiniMax-M3 72.7%
  59. 59 Gemini 3.5 Flash-Lite 72.6%
  60. 60 Gemini 3 Pro 72.6%
  61. 61 GLM-4.5 72.4%
  62. 62 GLM-5.3 72.4%
  63. 63 DeepSeek-R1 May 2025 72.3%
  64. 64 Qwen 3.5 Flash (hosted 35B-A3B) 72.2%
  65. 65 DeepSeek-V4-Pro 72.1%
  66. 66 Claude Sonnet 4 thinking 72%
  67. 67 GPT-4.1 mini 71.9%
  68. 68 Claude Opus 4 non-thinking 71.9%
  69. 69 MiMo-V2.5 71.8%
  70. 70 Gemini 3.1 Flash-Lite 71.8%
  71. 71 Kimi K2 Thinking 71.7%
  72. 72 Grok 4.5 71.7%
  73. 73 gpt-oss-120b 71.6%
  74. 74 Qwen3.6 27B 71.3%
  75. 75 GPT-5.4 Mini 71.2%
  76. 76 GLM-5.1 71.2%
  77. 77 Gemini 2.5 Flash Apr 2025, non-thinking 71.2%
  78. 78 DeepSeek-V3 Mar 2025 71.1%
  79. 79 Grok 4.6 71.1%
  80. 80 Qwen 3.8 27B 70.8%
  81. 81 Grok 4.3 70.8%
  82. 82 DeepSeek V4 Flash 0731 70.7%
  83. 83 Qwen3-235B-A22B 70.7%
  84. 84 Ling-3.0-flash 70.7%
  85. 85 Mistral Medium 3.1 70.3%
  86. 86 Kimi K2 Jul 2025 70.2%
  87. 87 Claude 3.5 Sonnet 70.2%
  88. 88 GLM-5 thinking 70%
  89. 89 Gemini 2.0 Flash Thinking Jan 2025 69.8%
  90. 90 o3-mini 69.4%
  91. 91 GLM-4.7 68.8%
  92. 92 Gemini 2.0 Pro 68.2%
  93. 93 DeepSeek-V3.2 thinking 68.2%
  94. 94 MiniMax-M2.5 68.2%
  95. 95 Mistral Medium 3.5 68%
  96. 96 Gemini 2.0 Flash Feb 2025 67.7%
  97. 97 Claude Haiku 4.5 thinking 67.5%
  98. 98 GPT-5.4 Nano 67.4%
  99. 99 GPT-5 nano 67.4%
  100. 100 Grok-2 Dec 2024 67%
  101. 101 Llama 4 Maverick 66.6%
  102. 102 MiniMax-M2.7 66.6%
  103. 103 MiniMax-M2.1 66.3%
  104. 104 Gemini 2.5 Flash-Lite Sep 2025, non-thinking 66.2%
  105. 105 GLM-4.6 66.2%
  106. 106 Grok 4 65.1%
  107. 107 Mistral Large 2 Nov 2024 63.8%
  108. 108 Llama 3.3 Nemotron Super (Thinking) 63.7%
  109. 109 gpt-oss-20b 63.7%
  110. 110 Magistral Medium 1.2 61.9%
  111. 111 Cohere Command A 61.4%
  112. 112 Llama 3.1-405B 60.9%
  113. 113 Jamba 1.6 Large 60.9%
  114. 114 GPT-4.1 nano 60.8%
  115. 115 GPT-4o mini 60.5%
  116. 116 Magistral Small 1.2 60.3%
  117. 117 Llama 3.3 Nemotron Super (non-thinking) 60.2%
  118. 118 Gemini 1.5 Pro Sep 2024 59.5%
  119. 119 Llama 3.3 70B 59.4%
  120. 120 Laguna XS.2 59%
  121. 121 Mistral Small 3.1 58.3%
  122. 122 Jamba 1.5 Large 58.2%
  123. 123 Claude 3.5 Haiku 57.4%
  124. 124 Llama 3.1-70B 56.2%
  125. 125 Llama 4 Scout 55.2%
  126. 126 mistral-small-2402 49.1%
  127. 127 Gemini 1.5 Flash Sep 2024 48.2%
  128. 128 Jamba 1.6 Mini 44.6%
  129. 129 Jamba 1.5 Mini 41.9%
  130. 130 Llama 3.1-8B 32.3%
  131. 131 Laguna M.1 1.64%
#EntryScoreSettingDateSource
1 Muse Spark 1.2 80.4% Aug 2026 Official leaderboard
2 Muse Spark 1.1 79.7% Jul 2026 Official leaderboard
3 Muse Spark 77.7% Apr 2026 Official leaderboard
4 Claude Sonnet 4.6 77.1% Feb 2026 Official leaderboard
5 Claude Fable 5 76.9% Jun 2026 Official leaderboard
6 GPT-5.6 Luna 76.2% Jul 2026 Official leaderboard
7 GPT-5.6 Terra 76.2% Jul 2026 Official leaderboard
8 Claude Opus 4.6 76% thinking Feb 2026 Official leaderboard
9 Claude Fable 5.1 76% Sep 2026 Official leaderboard
10 Grok 3 75.9% Apr 2025 Official leaderboard
11 GPT-5.2 75.8% Dec 2025 Official leaderboard
12 Kimi K3 75.7% Jul 2026 Official leaderboard
13 Grok 4 Fast 75.7% reasoning Sep 2025 Official leaderboard
14 Claude Opus 4.8 75.6% May 2026 Official leaderboard
15 Claude Sonnet 5 75.6% Jun 2026 Official leaderboard
16 GLM-5.3-Flash 75.6% Aug 2026 Official leaderboard
17 Qwen 3.8 Max 75.5% Aug 2026 Official leaderboard
18 Inkling-Small 75.5% Jul 2026 Official leaderboard
19 Qwen3.7-Max 75.3% May 2026 Official leaderboard
20 Inkling 75.3% Jul 2026 Official leaderboard
21 Claude Opus 4.7 75.3% Apr 2026 Official leaderboard
22 GPT-5 mini 75.2% Aug 2025 Official leaderboard
23 Claude Opus 5 75.1% Jul 2026 Official leaderboard
24 GPT-4.1 75.1% Apr 2025 Official leaderboard
25 GPT-5.5 75% Apr 2026 Official leaderboard
26 GPT-5.1 74.9% Nov 2025 Official leaderboard
27 Claude Opus 4.5 74.9% thinking Nov 2025 Official leaderboard
28 Gemini 3.6 Flash 74.9% Jul 2026 Official leaderboard
29 o4-mini 74.8% Apr 2025 Official leaderboard
30 GPT-5.6 Sol 74.8% Jul 2026 Official leaderboard
31 Qwen 3.6 Plus 74.7% Mar 2026 Official leaderboard
32 Gemini 3.7 Flash 74.7% Aug 2026 Official leaderboard
33 Kimi K2.6 74.7% Apr 2026 Official leaderboard
34 o3 74.6% Apr 2025 Official leaderboard
35 GPT-4o Aug 2024 74.5% Aug 2024 Official leaderboard
36 Gemini 3.8 Flash 74.5% Sep 2026 Official leaderboard
37 Gemini 3.5 Flash 74.4% May 2026 Official leaderboard
38 o1 74.3% Dec 2024 Official leaderboard
39 Kimi K2.5 74.2% thinking Jan 2026 Official leaderboard
40 Grok 4.20 74.1% reasoning Feb 2026 Official leaderboard
41 Claude 3.7 Sonnet 74% thinking Feb 2025 Official leaderboard
42 Qwen3-Max 74% Sep 2025 Official leaderboard
43 GPT-5.4 74% xhigh Mar 2026 Official leaderboard
44 Gemini 3 Flash 73.9% Dec 2025 Official leaderboard
45 MiMo-V2.5-Pro 73.8% Apr 2026 Official leaderboard
46 Claude Opus 4.1 73.7% thinking Aug 2025 Official leaderboard
47 GPT-5 73.4% Aug 2025 Official leaderboard
48 GLM-5.2 73.3% Jun 2026 Official leaderboard
49 Claude Sonnet 4.5 73.3% thinking Sep 2025 Official leaderboard
50 Grok 4.1 Fast 73.1% reasoning Nov 2025 Official leaderboard
51 Nemotron 3 Ultra 73.1% Jun 2026 Official leaderboard
52 Mistral Large 3 73.1% Dec 2025 Official leaderboard
53 DeepSeek V4 Pro 0813 73.1% Aug 2026 Official leaderboard
54 Grok-3 mini 73% reasoning Apr 2025 Official leaderboard
55 Gemini 2.5 Pro Jun 2025 72.9% Jun 2025 Official leaderboard
56 Gemini 3.1 Pro 72.9% Feb 2026 Official leaderboard
57 Gemini 2.5 Flash Sep 2025 72.7% non-thinking Sep 2025 Official leaderboard
58 MiniMax-M3 72.7% Jun 2026 Official leaderboard
59 Gemini 3.5 Flash-Lite 72.6% Jul 2026 Official leaderboard
60 Gemini 3 Pro 72.6% Nov 2025 Official leaderboard
61 GLM-4.5 72.4% Aug 2025 Official leaderboard
62 GLM-5.3 72.4% Aug 2026 Official leaderboard
63 DeepSeek-R1 May 2025 72.3% May 2025 Official leaderboard
64 Qwen 3.5 Flash (hosted 35B-A3B) 72.2% Feb 2026 Official leaderboard
65 DeepSeek-V4-Pro 72.1% Apr 2026 Official leaderboard
66 Claude Sonnet 4 72% thinking May 2025 Official leaderboard
67 GPT-4.1 mini 71.9% Apr 2025 Official leaderboard
68 Claude Opus 4 71.9% non-thinking May 2025 Official leaderboard
69 MiMo-V2.5 71.8% Apr 2026 Official leaderboard
70 Gemini 3.1 Flash-Lite 71.8% Mar 2026 Official leaderboard
71 Kimi K2 Thinking 71.7% Nov 2025 Official leaderboard
72 Grok 4.5 71.7% Jul 2026 Official leaderboard
73 gpt-oss-120b 71.6% Aug 2025 Official leaderboard
74 Qwen3.6 27B 71.3% Apr 2026 Official leaderboard
75 GPT-5.4 Mini 71.2% Mar 2026 Official leaderboard
76 GLM-5.1 71.2% Apr 2026 Official leaderboard
77 Gemini 2.5 Flash Apr 2025 71.2% non-thinking Apr 2025 Official leaderboard
78 DeepSeek-V3 Mar 2025 71.1% Mar 2025 Official leaderboard
79 Grok 4.6 71.1% Aug 2026 Official leaderboard
80 Qwen 3.8 27B 70.8% Aug 2026 Official leaderboard
81 Grok 4.3 70.8% Apr 2026 Official leaderboard
82 DeepSeek V4 Flash 0731 70.7% Jul 2026 Official leaderboard
83 Qwen3-235B-A22B 70.7% Apr 2025 Official leaderboard
84 Ling-3.0-flash 70.7% Aug 2026 Official leaderboard
85 Mistral Medium 3.1 70.3% Aug 2025 Official leaderboard
86 Kimi K2 Jul 2025 70.2% Jul 2025 Official leaderboard
87 Claude 3.5 Sonnet 70.2% Jun 2024 Official leaderboard
88 GLM-5 70% thinking Feb 2026 Official leaderboard
89 Gemini 2.0 Flash Thinking Jan 2025 69.8% Jan 2025 Official leaderboard
90 o3-mini 69.4% Jan 2025 Official leaderboard
91 GLM-4.7 68.8% Dec 2025 Official leaderboard
92 Gemini 2.0 Pro 68.2% Feb 2025 Official leaderboard
93 DeepSeek-V3.2 68.2% thinking Dec 2025 Official leaderboard
94 MiniMax-M2.5 68.2% Feb 2026 Official leaderboard
95 Mistral Medium 3.5 68% Apr 2026 Official leaderboard
96 Gemini 2.0 Flash Feb 2025 67.7% Feb 2025 Official leaderboard
97 Claude Haiku 4.5 67.5% thinking Oct 2025 Official leaderboard
98 GPT-5.4 Nano 67.4% Mar 2026 Official leaderboard
99 GPT-5 nano 67.4% Aug 2025 Official leaderboard
100 Grok-2 Dec 2024 67% Dec 2024 Official leaderboard
101 Llama 4 Maverick 66.6% Apr 2025 Official leaderboard
102 MiniMax-M2.7 66.6% Mar 2026 Official leaderboard
103 MiniMax-M2.1 66.3% Dec 2025 Official leaderboard
104 Gemini 2.5 Flash-Lite Sep 2025 66.2% non-thinking Sep 2025 Official leaderboard
105 GLM-4.6 66.2% Sep 2025 Official leaderboard
106 Grok 4 65.1% Jul 2025 Official leaderboard
107 Mistral Large 2 Nov 2024 63.8% Nov 2024 Official leaderboard
108 Llama 3.3 Nemotron Super (Thinking) 63.7% Official leaderboard
109 gpt-oss-20b 63.7% Aug 2025 Official leaderboard
110 Magistral Medium 1.2 61.9% Sep 2025 Official leaderboard
111 Cohere Command A 61.4% Mar 2025 Official leaderboard
112 Llama 3.1-405B 60.9% Jul 2024 Official leaderboard
113 Jamba 1.6 Large 60.9% Mar 2025 Official leaderboard
114 GPT-4.1 nano 60.8% Apr 2025 Official leaderboard
115 GPT-4o mini 60.5% Jul 2024 Official leaderboard
116 Magistral Small 1.2 60.3% Sep 2025 Official leaderboard
117 Llama 3.3 Nemotron Super (non-thinking) 60.2% Official leaderboard
118 Gemini 1.5 Pro Sep 2024 59.5% Sep 2024 Official leaderboard
119 Llama 3.3 70B 59.4% Dec 2024 Official leaderboard
120 Laguna XS.2 59% Apr 2026 Official leaderboard
121 Mistral Small 3.1 58.3% Mar 2025 Official leaderboard
122 Jamba 1.5 Large 58.2% Aug 2024 Official leaderboard
123 Claude 3.5 Haiku 57.4% Oct 2024 Official leaderboard
124 Llama 3.1-70B 56.2% Jul 2024 Official leaderboard
125 Llama 4 Scout 55.2% Apr 2025 Official leaderboard
126 mistral-small-2402 49.1% Feb 2024 Official leaderboard
127 Gemini 1.5 Flash Sep 2024 48.2% Aug 2024 Official leaderboard
128 Jamba 1.6 Mini 44.6% Mar 2025 Official leaderboard
129 Jamba 1.5 Mini 41.9% Aug 2024 Official leaderboard
130 Llama 3.1-8B 32.3% Jul 2024 Official leaderboard
131 Laguna M.1 1.64% Apr 2026 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Grok 4 Fast
non-reasoni… reasoning
75.7%
Claude Opus 4.5
non-thinking
74.9%
Claude 3.7 Sonnet
non-thinking thinking
74%
Claude Opus 4.1
non-thinking thinking
73.7%
Grok 4.1 Fast
non-reasoni… reasoning
73.1%
Gemini 2.5 Flash
thinking
72.7%
Claude Sonnet 4
non-thinking thinking
72%
Gemini 2.5 Flash
thinking
71.2%
Gemini 2.5 Flash-Lite
thinking non-thinking
66.2%
ModelEffortHarnessScoreCost per taskOutput tokens
Muse Spark 1.2 80.4%
Muse Spark 1.1 79.7%
Muse Spark 77.7%
Claude Sonnet 4.6 77.1%
Claude Fable 5 76.9%
GPT-5.6 Luna 76.2%
GPT-5.6 Terra 76.2%
Claude Opus 4.6 thinking 76%
Claude Fable 5.1 76%
Grok 3 75.9%
GPT-5.2 75.8%
Kimi K3 75.7%
Grok 4 Fast reasoning 75.7%
Claude Opus 4.8 75.6%
Claude Sonnet 5 75.6%
GLM-5.3-Flash 75.6%
Qwen 3.8 Max 75.5%
Inkling-Small 75.5%
Qwen3.7-Max 75.3%
Inkling 75.3%
Claude Opus 4.7 75.3%
GPT-5 mini 75.2%
Claude Opus 5 75.1%
GPT-4.1 75.1%
GPT-5.5 75%
GPT-5.1 74.9%
Claude Opus 4.5 thinking 74.9%
Gemini 3.6 Flash 74.9%
o4-mini 74.8%
GPT-5.6 Sol 74.8%
Qwen 3.6 Plus 74.7%
Gemini 3.7 Flash 74.7%
Kimi K2.6 74.7%
o3 74.6%
GPT-4o Aug 2024 74.5%
Gemini 3.8 Flash 74.5%
Gemini 3.5 Flash 74.4%
Claude Opus 4.5 non-thinking 74.3%
o1 74.3%
Kimi K2.5 thinking 74.2%
Grok 4.20 reasoning 74.1%
Claude 3.7 Sonnet thinking 74%
Qwen3-Max 74%
GPT-5.4 xhigh 74%
Gemini 3 Flash 73.9%
MiMo-V2.5-Pro 73.8%
Claude Opus 4.1 thinking 73.7%
Qwen3-Max 73.5%
GPT-5 73.4%
GLM-5.2 73.3%
Claude Sonnet 4.5 thinking 73.3%
Grok 4.1 Fast reasoning 73.1%
Nemotron 3 Ultra 73.1%
Mistral Large 3 73.1%
DeepSeek V4 Pro 0813 73.1%
Grok-3 mini reasoning 73%
Gemini 2.5 Pro Jun 2025 72.9%
Gemini 3.1 Pro 72.9%
Gemini 2.5 Flash Sep 2025 non-thinking 72.7%
MiniMax-M3 72.7%
Gemini 3.5 Flash-Lite 72.6%
Gemini 3 Pro 72.6%
Claude 3.7 Sonnet non-thinking 72.4%
GLM-4.5 72.4%
Gemini 2.5 Flash Sep 2025 thinking 72.4%
GLM-5.3 72.4%
DeepSeek-R1 May 2025 72.3%
Qwen 3.5 Flash (hosted 35B-A3B) 72.2%
DeepSeek-V4-Pro 72.1%
Claude Sonnet 4 thinking 72%
GPT-4.1 mini 71.9%
Claude Opus 4 non-thinking 71.9%
MiMo-V2.5 71.8%
Gemini 3.1 Flash-Lite 71.8%
Kimi K2 Thinking 71.7%
Grok 4.5 71.7%
gpt-oss-120b 71.6%
Grok 4 Fast non-reasoning 71.6%
Claude Opus 4.1 non-thinking 71.5%
Qwen3.6 27B 71.3%
GPT-5.4 Mini 71.2%
GLM-5.1 71.2%
Gemini 2.5 Flash Apr 2025 non-thinking 71.2%
GPT-4o Aug 2024 71.1%
Grok-3 mini reasoning 71.1%
DeepSeek-V3 Mar 2025 71.1%
Grok 4.6 71.1%
Qwen 3.8 27B 70.8%
Grok 4.3 70.8%
DeepSeek V4 Flash 0731 70.7%
Qwen3-235B-A22B 70.7%
Ling-3.0-flash 70.7%
Gemini 2.5 Flash Apr 2025 thinking 70.5%
Mistral Medium 3.1 70.3%
Kimi K2 Jul 2025 70.2%
Claude 3.5 Sonnet 70.2%
GLM-5 thinking 70%
Gemini 2.0 Flash Thinking Jan 2025 69.8%
Claude Sonnet 4 non-thinking 69.6%
o3-mini 69.4%
GLM-4.7 68.8%
Gemini 2.0 Pro 68.2%
DeepSeek-V3.2 thinking 68.2%
MiniMax-M2.5 68.2%
Mistral Medium 3.5 68%
DeepSeek-V3 Mar 2025 67.9%
Gemini 2.0 Flash Feb 2025 67.7%
Claude Haiku 4.5 thinking 67.5%
GPT-5.4 Nano 67.4%
GPT-5 nano 67.4%
Grok-2 Dec 2024 67%
Llama 4 Maverick 66.6%
MiniMax-M2.7 66.6%
MiniMax-M2.1 66.3%
Gemini 2.5 Flash-Lite Sep 2025 non-thinking 66.2%
GLM-4.6 66.2%
Gemini 2.0 Flash Feb 2025 65.3%
Grok 4 65.1%
Gemini 2.5 Flash-Lite Sep 2025 thinking 64.7%
Mistral Large 2 Nov 2024 63.8%
gpt-oss-20b 63.7%
Magistral Medium 1.2 61.9%
Grok 4.1 Fast non-reasoning 61.9%
Cohere Command A 61.4%
Llama 3.1-405B 60.9%
Jamba 1.6 Large 60.9%
GPT-4.1 nano 60.8%
GPT-4o mini 60.5%
Magistral Small 1.2 60.3%
Gemini 1.5 Pro Sep 2024 59.5%
Llama 3.3 70B 59.4%
Laguna XS.2 59%
Mistral Small 3.1 58.3%
Jamba 1.5 Large 58.2%
Claude 3.5 Haiku 57.4%
Llama 3.1-70B 56.2%
Llama 4 Scout 55.2%
mistral-small-2402 49.1%
Gemini 1.5 Flash Sep 2024 48.2%
Jamba 1.6 Mini 44.6%
Jamba 1.5 Mini 41.9%
Llama 3.1-8B 32.3%
Laguna M.1 1.64%

9 models have held number one since 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

Other labs Anthropic OpenAI xAI Meta
Own leaderboard Not tested since Sep 2026
Muse Spark 1.2 80.4%

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Muse Spark 1.2 Claude Sonnet 4.6 Grok 3 GPT-4o Claude 3.5 Sonnet mistral-small-2402 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Muse Spark 1.2 GPT-4o

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Law, finance and medicine

All 140 in Law, finance and medicine
  • HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
  • MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy. Qwen 3.8 Max 80.4%
  • HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • KBL In use Models answer Korean legal knowledge and reasoning tasks and Korean bar exam questions. It has 7 knowledge tasks, 4 reasoning tasks and 2,510 bar exam items. Solar Pro 4 77.5%
  • KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. Solar Pro 4 93.2%
  • PRBench In use Model answers 1,100 expert-written law and finance questions. LLM judge grades against weighted expert rubrics of 10-30 criteria each. Muse Spark 1.3 61.6%