Benchmarks / Coding

Terminal-Bench v4-AA

AA's 66-task Terminal-Bench v4.0, three repeats, average pass@1. Distinct from the upstream leaderboard and v2.1.

145models
–Open Index weight
–last updated

Terminal-Bench v4-AA

Higher is better · best effort per model

Chart settings
145 of 145 selected

full table below

#ModelEffortScoreCost / taskHarness
1Sonnet 5.5max63.6%–mini-SWE-agent
2Opus 5.5max59.6%–mini-SWE-agent
3GPT-6 Astraxhigh59.6%–mini-SWE-agent
4Muse Wonder Full (Max)max57.1%–mini-SWE-agent
5Gemini 4high57.1%–mini-SWE-agent
6GPT-6.1 Solmax56.1%–mini-SWE-agent
7Fable 5.1xhigh55.1%–mini-SWE-agent
8Muse Wondermax52.0%–mini-SWE-agent
9Opus 5max49.0%–mini-SWE-agent
10GPT-6 Solmax43.9%–mini-SWE-agent
11Fable 5max42.4%–mini-SWE-agent
12GLM 5.3max41.9%–mini-SWE-agent
13GPT-5.6 Solmax39.9%–mini-SWE-agent
14Qwen 3.8 Maxmax38.9%–mini-SWE-agent
15GPT-5.6 Terramax35.4%–mini-SWE-agent
16MiMo V2.6 Pro–34.8%–mini-SWE-agent
17Muse Spark 1.3max33.3%–mini-SWE-agent
18Step 5 Preview–33.3%–mini-SWE-agent
19Ling 3.1 Flash–33.3%–mini-SWE-agent
20Haiku 5.5max32.8%–mini-SWE-agent
21GLM 5.3 Flashmax32.8%–mini-SWE-agent
22DeepSeek V4.1 Flashmax26.8%–mini-SWE-agent
23Mistral Large 4–26.8%–mini-SWE-agent
24Grok 4.7xhigh25.8%–mini-SWE-agent
25Qwen 3.8 Flash–25.3%–mini-SWE-agent
26MiMo V2.6 Flash–22.7%–mini-SWE-agent
27Opus 4.8max21.7%–mini-SWE-agent
28Grok 4.6high21.2%–mini-SWE-agent
29Gemini 3.8 Flashhigh19.7%–mini-SWE-agent
30GPT-5.5xhigh14.6%–mini-SWE-agent
31DeepSeek V4 Pro 0424max14.6%–mini-SWE-agent
32Sonnet 5max14.1%–mini-SWE-agent
33DeepSeek V4 Pro 0813max14.1%–mini-SWE-agent
34Gemini 3.7 Flashhigh13.6%–mini-SWE-agent
35Kimi K3max12.6%–mini-SWE-agent
36GPT-6 Lunamax12.6%–mini-SWE-agent
37GPT-5.5 Instant–12.6%–mini-SWE-agent
38DeepSeek V4 Flashmax12.1%–mini-SWE-agent
39DeepSeek V4 Flash 0731max12.1%–mini-SWE-agent
40GPT-5.6 Lunamax11.6%–mini-SWE-agent
41Qwen 3.8 2.4T-A95B–11.1%–mini-SWE-agent
42Grok 4.5high10.6%–mini-SWE-agent
43Muse Spark Open (Xhigh)xhigh8.1%–mini-SWE-agent
44Muse Spark 1.2xhigh7.1%–mini-SWE-agent
45Gemini 3.6 Flashhigh7.1%–mini-SWE-agent
46Gemini 3.5 Flashhigh6.6%–mini-SWE-agent
47Muse Spark 1.1xhigh6.1%–mini-SWE-agent
48Qwen 3.8 27Bxhigh5.6%–mini-SWE-agent
49Gemini 3.1 Pro–4.0%–mini-SWE-agent
50JT-4.1 Flash 236B A21B (Reasoning)–3.0%–mini-SWE-agent
51Beam 1.0–3.0%–mini-SWE-agent
52Sonnet 4.6max3.0%–mini-SWE-agent
53DeepSeek V4 Flash 0420high3.0%–mini-SWE-agent
54MiniMax M3–2.0%–mini-SWE-agent
55GLM 5.1–2.0%–mini-SWE-agent
56GPT-5.4 Minixhigh2.0%–mini-SWE-agent
57K2 Horizon 375B A23B–1.5%–mini-SWE-agent
58Qwen 3.7 Maxmax1.5%–mini-SWE-agent
59GLM 5.2max1.0%–mini-SWE-agent
60Quasar 438B (Max, Based on GLM-5.2)max1.0%–mini-SWE-agent
61Kimi K2.7 Code–1.0%–mini-SWE-agent
62Inkling Small–1.0%–mini-SWE-agent
63Qwen 3.7 Plus–1.0%–mini-SWE-agent
64Inklingxhigh1.0%–mini-SWE-agent
65Solar Mini 4–1.0%–mini-SWE-agent
66Gemini 3.5 Flash-Litehigh1.0%–mini-SWE-agent
67K2 Horizon 7B–1.0%–mini-SWE-agent
68Kimi K2.6–0.5%–mini-SWE-agent
69Hy3–0.5%–mini-SWE-agent
70Nemotron 3 Ultramax0.5%–mini-SWE-agent
71GPT-5.4 Nanoxhigh0.5%–mini-SWE-agent
72Muse Glimmer (High)high0.5%–mini-SWE-agent
73Ring-2.6-1Txhigh0.5%–mini-SWE-agent
74Gemini 3.1 Flash-Litehigh0.5%–mini-SWE-agent
75Command A–0.5%–mini-SWE-agent
76Nemotron 3.5 Lightning–0.5%–mini-SWE-agent
77Qwen 3.5 9B–0.5%–mini-SWE-agent
78Trinity Large Thinking–0.5%–mini-SWE-agent
79North Mini Code–0.5%–mini-SWE-agent
80Solar Pro 4–0.0%–mini-SWE-agent
81Nemotron 3.5 Super VL–0.0%–mini-SWE-agent
82Apodex 1.1–0.0%–mini-SWE-agent
83MiMo V2.5 Pro–0.0%–mini-SWE-agent
84K2 Horizon MoVA 36B A4B–0.0%–mini-SWE-agent
85Grok 4.3high0.0%–mini-SWE-agent
86Ling 3.0 Flash–0.0%–mini-SWE-agent
87MiniMax M2.7–0.0%–mini-SWE-agent
88Ling 3.0 Flash Fin–0.0%–mini-SWE-agent
89MiMo V2.5–0.0%–mini-SWE-agent
90A.X-K2–0.0%–mini-SWE-agent
91Qwen 3.6 27B–0.0%–mini-SWE-agent
92Nemotron 3.5 Bolt 74B-A9B–0.0%–mini-SWE-agent
93Sonnet 4.5–0.0%–mini-SWE-agent
94Ling 3.0 Flash–0.0%–mini-SWE-agent
95LongCat 2.0–0.0%–mini-SWE-agent
96Qwen 3.5 397B-A17B–0.0%–mini-SWE-agent
97Qwen 3.6 35B-A3B–0.0%–mini-SWE-agent
98Haiku 4.5–0.0%–mini-SWE-agent
99GPT-5 Minihigh0.0%–mini-SWE-agent
100Gemini 2.5 Pro–0.0%–mini-SWE-agent
101K2 Horizon 3.7B–0.0%–mini-SWE-agent
102Qwen 3.5 122B-A10B–0.0%–mini-SWE-agent
103DeepSeek V3.1–0.0%–mini-SWE-agent
104Gemma 4 31B–0.0%–mini-SWE-agent
105Mistral Medium 3.5high0.0%–mini-SWE-agent
106Nemotron 3 Super 120B A12B (Reasoning)–0.0%–mini-SWE-agent
107Qwen 3 235B-A22B–0.0%–mini-SWE-agent
108Granite 4.2 30B–0.0%–mini-SWE-agent
109MiniCPM5-2B–0.0%–mini-SWE-agent
110Mercury 2.5–0.0%–mini-SWE-agent
111HyperNova 60B 2605 (High, Based on gpt-oss-120b)high0.0%–mini-SWE-agent
112gpt-oss-120bhigh0.0%–mini-SWE-agent
113DeepSeek R1–0.0%–mini-SWE-agent
114Mistral Small 4–0.0%–mini-SWE-agent
115Granite 4.2 8B–0.0%–mini-SWE-agent
116Ling 3.0 Tiny–0.0%–mini-SWE-agent
117o3 minihigh0.0%–mini-SWE-agent
118Mercury 2high0.0%–mini-SWE-agent
119Qwen 3 30B-A3B–0.0%–mini-SWE-agent
120DeepSeek V3none0.0%–mini-SWE-agent
121Llama 4 Mavericknone0.0%–mini-SWE-agent
122Mistral Large 3none0.0%–mini-SWE-agent
123Qwen 3 Codernone0.0%–mini-SWE-agent
124Mistral Medium 3.1none0.0%–mini-SWE-agent
125Granite 4.2 3B–0.0%–mini-SWE-agent
126gpt-oss-20bhigh0.0%–mini-SWE-agent
127Gemma 4 E4B (Reasoning)–0.0%–mini-SWE-agent
128NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)–0.0%–mini-SWE-agent
129Devstral 2none0.0%–mini-SWE-agent
130Solar Pro 3high0.0%–mini-SWE-agent
131Devstral Small 2none0.0%–mini-SWE-agent
132Qwen 3 32B–0.0%–mini-SWE-agent
133Mistral Small 3.1none0.0%–mini-SWE-agent
134Mistral Small 3.2none0.0%–mini-SWE-agent
135Llama 4 Scoutnone0.0%–mini-SWE-agent
136Qwen 3 14B–0.0%–mini-SWE-agent
137Celeris-1none0.0%–mini-SWE-agent
138Ministral 3 14Bnone0.0%–mini-SWE-agent
139Ministral 3 8Bnone0.0%–mini-SWE-agent
140Qwen 3 8B–0.0%–mini-SWE-agent
141Gemma 3 27Bnone0.0%–mini-SWE-agent
142Ministral 3 3Bnone0.0%–mini-SWE-agent
143NVIDIA Nemotron 3 Nano 4B–0.0%–mini-SWE-agent
144Gemma 3 12Bnone0.0%–mini-SWE-agent
145K2 Horizon 0.9B–0.0%–mini-SWE-agent