Terminal-Bench 4.0 leaderboard

Frontier-difficulty tasks an agent completes in a real terminal: building, configuring, debugging and running software. Claude Opus 5.5 leads with 65.15%.

Models
52
Leader
Claude Opus 5.5
Source
Vals AI
As of
Oct 24, 2026
  1. 1Claude Opus 5.565.2%
  2. 2Claude Sonnet 5.564.1%
  3. 3Kimi K3.1 Pro63.6%
  4. 4GPT-6 Astra59.6%
  5. 5Claude Fable 5.158.1%
  6. 6Muse Wonder Full58.1%
  7. 7Gemini 4 Argon57.6%
  8. 8GPT-6.1 Sol55.0%
  9. 9Claude Opus 553.5%
  10. 10Muse Wonder49.5%
Terminal-Bench 4.0, accuracy against cost per test
Open weightsProprietaryBest for the price
0%18%35%53%70%$.1$1$10Cost per test, log scale →Claude Opus 5.5Kimi K3.1 ProGPT-6.1 SolGLM-5.3-FlashMiMo-V2.6-ProMiMo-V2.6-Flash

Full leaderboard

RankModelLabAccuracyCost / testHarnessPrice in / out
1Claude Opus 5.5Anthropic65.15%$13.20Mini-SWE-agent$4 / $20
2Claude Sonnet 5.5Anthropic64.14%$16.51Mini-SWE-agent$2 / $10
3Kimi K3.1 ProMoonshot AI63.64%$4.57Mini-SWE-agent$2 / $10
4GPT-6 AstraOpenAI59.60%$9.58Mini-SWE-agent$10 / $50
5Claude Fable 5.1Anthropic58.08%$17.18Mini-SWE-agent$10 / $50
6Muse Wonder FullMeta58.08%$10.53Mini-SWE-agent—
7Gemini 4 ArgonGoogle57.58%$17.64Mini-SWE-agent$2 / $10
8GPT-6.1 SolOpenAI55.05%$1.72Mini-SWE-agent$2 / $10
9Claude Opus 5Anthropic53.53%$18.60Mini-SWE-agent$5 / $25
10Muse WonderMeta49.49%$14.66Mini-SWE-agent$4 / $20
11GPT-6 SolOpenAI44.44%$5.79Mini-SWE-agent$2 / $10
12Claude Fable 5Anthropic41.41%$30.34Mini-SWE-agent$10 / $50
13GLM-5.3Z.ai38.89%$9.37Mini-SWE-agent$0.079 / $1.40
15GPT-5.6 SolOpenAI37.88%$7.98Mini-SWE-agent$4 / $20
16GLM-5.3-FlashZ.ai37.37%$1.22Mini-SWE-agent$0.15 / $0.50
17Claude Haiku 5.5Anthropic36.36%$3.12Mini-SWE-agent$0.10 / $0.50
18Qwen3.8-MaxAlibaba34.34%$10.63Mini-SWE-agent$2 / $6
20Step 5 PreviewStepFun31.82%$3.11Mini-SWE-agent$1 / $2.70
21MiMo-V2.6-ProXiaomi31.31%$0.50Mini-SWE-agent$0.43 / $0.87
22Grok 4.7xAI28.79%$18.01Mini-SWE-agent$2 / $6
23Mistral Large 4Mistral AI27.78%$10.25Mini-SWE-agent$1.36 / $4.18
24MiniMax-M3.1-FlashMiniMax25.25%$0.73Mini-SWE-agent$0.15 / $0.60
25Muse Spark 1.3 MaxMeta24.75%$6.65Mini-SWE-agent—
26MiMo-V2.6-FlashXiaomi24.24%$0.22Mini-SWE-agent$0.14 / $0.28
27Claude Opus 4.8Anthropic23.23%$17.14Mini-SWE-agent$5 / $25
28GPT-5.6 TerraOpenAI22.73%$5.60Mini-SWE-agent$2 / $12
29Ember-1Fireworks AI20.71%$19.48Mini-SWE-agent$3 / $15
30DeepSeek-V4.1-Flash-20260910DeepSeek19.70%$0.50Mini-SWE-agent$0.15 / $0.60
31Gemini 3.8 FlashGoogle19.70%$8.65Mini-SWE-agent$0.82 / $4.13
32DeepSeek-V4-FlashDeepSeek18.69%$0.77Mini-SWE-agent$0.042 / $0.18
33Grok 4.6xAI17.17%$5.06Mini-SWE-agent$2 / $6
34Kimi K3Moonshot AI17.17%$12.48Mini-SWE-agent$0.24 / $11
36GPT-6 LunaOpenAI13.64%$0.35Mini-SWE-agent$0.10 / $0.50
37Gemini 3.7 FlashGoogle12.12%$8.71Mini-SWE-agent$0.75 / $3.75
38GPT-5.6 LunaOpenAI11.62%$0.73Mini-SWE-agent$0.20 / $1.20
39DeepSeek-V4-ProDeepSeek11.11%$3.45Mini-SWE-agent$0.66 / $1.98
40Claude Sonnet 5Anthropic10.61%$26.05Mini-SWE-agent$2 / $10
41Muse Spark 1.3Meta10.61%$7.04Mini-SWE-agent$1.25 / $4.25
42Grok 4.5xAI8.59%$5.52Mini-SWE-agent$2 / $6
43Hy4 previewTencent8.08%$2.10Mini-SWE-agent$0.75 / $2.25
44Muse Spark OpenMeta6.57%$7.26Mini-SWE-agent$1.25 / $4.25
45Gemini 3.5 FlashGoogle6.06%$5.98Mini-SWE-agent$1.50 / $9
46Muse Spark 1.2Meta6.06%$4.14Mini-SWE-agent$1.25 / $4.25
47Beam 1.0Reflection AI4.04%N/AMini-SWE-agent—
48Mercury 3Inception3.03%$0.42Mini-SWE-agent$0.10 / $0.35
49Gemini 3.1 Pro PreviewGoogle2.52%$4.11Mini-SWE-agent$2 / $12
50GPT-5.4 miniOpenAI2.52%$1.44Mini-SWE-agent$0.75 / $4.50
51MiniMax-M3MiniMax1.01%$7.23Mini-SWE-agent$0.30 / $1.20
52InklingThinky0.51%$2.50Mini-SWE-agent$1 / $4.05
53MiMo-V2.5-ProXiaomi0.51%$0.23Mini-SWE-agent$0.30 / $0.61
54Inkling-SmallThinky0.00%$0.32Mini-SWE-agent$0.45 / $1.20
55Mercury 2.5Inception0.00%$0.37Mini-SWE-agent$0.04 / $0.15

Related leaderboards

Sources

  • Vals AI, as of Oct 24, 2026
  • Provider list prices, as of Oct 24, 2026

Every number on this page comes from these sources; models without a published number show none. Download the data. Cite as “Nexus Explorer, nexus.inc/models”.