Benchmarks / Coding

SciCode

SciCode: implementing scientific research code from paper descriptions; sub-problem accuracy.

153models
–Open Index weight
–last updated

SciCode

Higher is better · best effort per model

Chart settings
153 of 153 selected

full table below

#ModelEffortScoreCost / taskHarness
1Opus 5.5max66.9%$0.469–
2Fable 5.1max63.1%$0.659–
3Muse Wonder Full (Max)max62.6%––
4Gemini 4high61.8%$0.106–
5Sonnet 5.5max61.0%$0.278–
6Fable 5max61.0%$0.268–
7MiMo V2.6 Pro–60.9%$0.0078–
8Muse Wondermax60.3%$0.135–
9Gemini 3.7 Flashmedium59.8%––
10Muse Spark 1.3xhigh59.7%$0.037–
11Kimi K3max59.5%$0.060–
12GLM 5.3max59.0%$0.082–
13Step 5 Preview–58.9%$0.020–
14Muse Spark 1.1xhigh58.8%$0.031–
15Gemini 3.1 Pro–58.7%$0.083–
16Muse Spark Open (Xhigh)xhigh58.4%$0.040–
17GPT-5.6 Solhigh57.8%$0.035–
18Grok 4.7high57.8%$0.099–
19GPT-6 Solmax57.6%$0.042–
20Muse Spark 1.2xhigh57.4%$0.060–
21Gemini 3.8 Flashhigh56.6%$0.077–
22GPT-6 Astramax56.5%$0.232–
23Grok 4.6high56.5%$0.027–
24Opus 5max56.4%$0.118–
25GPT-5.5high56.1%$0.107–
26GPT-6.1 Solhigh55.8%$0.013–
27Haiku 5.5max55.0%$0.011–
28GPT-5.6 Terramax55.0%$0.114–
29Grok 4.5high55.0%$0.026–
30GPT-6 Lunamax54.6%$0.0037–
31Opus 4.8max54.4%$0.266–
32Sonnet 5max54.3%$0.192–
33Mistral Large 4–54.2%$0.067–
34Ling 3.1 Flash–54.1%$0.015–
35Qwen 3.8 2.4T-A95B–54.1%$0.101–
36Gemini 3.5 Flashhigh53.9%$0.068–
37GPT-5.6 Lunamax53.6%$0.0093–
38Gemini 3.6 Flashhigh53.4%$0.030–
39Qwen 3.8 Maxmax53.2%$0.111–
40GPT-5.5 Instant–52.5%$0.025–
41GPT-5.4 Minixhigh52.1%$0.211–
42JT-4.1 Flash 236B A21B (Reasoning)–52.0%––
43DeepSeek V4.1 Flashmax51.9%$0.012–
44GLM 5.3 Flashmax51.6%$0.014–
45Kimi K2.6–51.5%$0.043–
46MiMo V2.6 Flash–51.3%$0.0028–
47GLM 5.2max51.2%$0.055–
48DeepSeek V4 Pro 0813max51.0%$0.050–
49DeepSeek V4 Pro 0424max50.8%$0.011–
50Qwen 3.8 Flash–50.6%$0.013–
51MiMo V2.5 Pro–50.6%$0.0049–
52DeepSeek V4 Flash 0731max50.3%$0.042–
53Sonnet 4.6max50.1%$0.150–
54MiniMax M2.7–50.1%––
55DeepSeek V4 Flashmax49.7%$0.034–
56Inkling Small–49.7%$0.023–
57Qwen 3.7 Maxmax49.5%$0.058–
58Agnes 2.5 Pro Beta–48.8%––
59Hy3–48.6%$0.0066–
60Beam 1.0–48.5%––
61Grok 4.3high48.3%$0.013–
62Quasar 438B (Max, Based on GLM-5.2)max48.1%$0.032–
63Solar Pro 4–48.0%$0.0088–
64Solar Open2 250B–48.0%––
65Nemotron 3.5 Super VL–47.8%$0.0044–
66Kimi K2.7 Code–47.8%$0.024–
67Solar Mini 4–47.6%$0.012–
68GPT-5.4 Nanoxhigh47.2%$0.023–
69MiniMax M3–47.1%$0.026–
70Inklingxhigh47.0%––
71Qwen 3.8 27Bxhigh46.6%$0.064–
72Gemini 2.5 Pro–46.3%$0.080–
73Qwen 3.7 Plus–46.1%$0.014–
74Sonnet 4.5–45.7%$0.129–
75Apodex 1.1–45.5%$0.021–
76Gemma 4 31B–45.5%$0.0000–
77DeepSeek V4 Flash 0420max45.3%$0.0033–
78Ring-2.6-1Txhigh45.0%$0.010–
79Muse Glimmer (High)high44.9%$0.0063–
80GLM 5.1–44.8%$0.044–
81Qwen 3.5 397B-A17B–44.8%$0.026–
82Ling 3.0 Flash–44.2%––
83MiMo V2.5–43.9%$0.0035–
84Step 3.7 Flashhigh43.9%––
85Gemini 3.1 Flash-Litehigh43.4%$0.018–
86Nex-N2-Pro–43.3%––
87K2 Horizon 375B A23B–42.9%––
88Qwen 3.6 27B–42.8%$0.020–
89o3 minihigh42.8%––
90Ling 3.0 Flash Fin–42.4%––
91Motif 3–42.2%––
92Haiku 4.5–42.2%$0.051–
93Ling 3.0 Flash–42.0%––
94K-EXAONE 2.0 0803–42.0%––
95Qwen 3 235B-A22B–41.4%––
96Gemini 3.5 Flash-Litehigh41.3%$0.013–
97A.X-K2–41.0%––
98Trinity Large Thinking–40.6%$0.012–
99Nemotron 3 Ultramax40.3%$0.011–
100Mistral Medium 3.5high40.2%$0.082–
101K2 Horizon MoVA 36B A4B–40.0%––
102Gemma 4 26B–40.0%––
103Nemotron 3.5 Bolt 74B-A9B–39.7%––
104Qwen 3.5 122B-A10B–39.7%$0.016–
105GPT-5 Minihigh39.0%$0.011–
106DeepSeek V3none39.0%$0.0010–
107gpt-oss-20bhigh38.9%$0.0019–
108Mistral Small 4–38.8%$0.0051–
109North Mini Code–38.8%$0.0000–
110Command A–38.5%$0.0000–
111Mercury 2.5–38.5%$0.0033–
112DeepSeek R1–38.3%$0.028–
113DeepSeek V3.1–38.0%––
114Granite 4.2 30B–37.8%$0.0060–
115Mercury 2high37.7%$0.0035–
116G9v3-39A5B–36.8%––
117Qwen 3.6 35B-A3B–36.6%$0.0095–
118Mistral Large 3none36.6%$0.0020–
119LongCat 2.0–36.3%$0.049–
120Nemotron 3 Super 120B A12B (Reasoning)–36.2%$0.0026–
121Qwen 3 Codernone36.2%$0.0031–
122Qwen 3 32B–36.0%––
123gpt-oss-120bhigh34.0%$0.0038–
124Qwen 3 30B-A3B–33.0%$0.0075–
125Devstral 2none32.8%––
126Mistral Medium 3.1none32.4%––
127Devstral Small 2none32.4%––
128Nemotron 3.5 Lightning–32.1%$0.0032–
129Llama 4 Mavericknone31.7%$0.0008–
130Granite 4.2 8B–31.5%$0.0025–
131Qwen 3 14B–30.7%––
132NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)–30.6%$0.0011–
133Qwen 3.5 9B–29.5%$0.0018–
134Mistral Small 3.2none28.6%$0.0003–
135Qwen 3 8B–27.9%––
136Mistral Small 3.1none27.8%$0.0003–
137K2 Horizon 7B–27.5%––
138MiniCPM5-2B–26.3%––
139Solar Pro 3high25.5%$0.0069–
140Granite 4.2 3B–25.3%$0.0016–
141Gemma 4 E4B (Reasoning)–24.4%$0.0003–
142Ling 3.0 Tiny–24.2%$0.0000–
143Ministral 3 14Bnone23.8%$0.0007–
144Gemma 3 27Bnone23.3%$0.0000–
145K2 Horizon 3.7B–22.0%––
146Celeris-1none21.6%$0.0017–
147Llama 4 Scoutnone21.3%$0.0009–
148Ministral 3 8Bnone20.7%$0.0006–
149Gemma 3 12Bnone16.4%$0.0000–
150NVIDIA Nemotron 3 Nano 4B–16.1%––
151Ministral 3 3Bnone15.3%$0.0005–
152LFM2.5-2.6B–14.4%––
153K2 Horizon 0.9B–6.9%––