# Vals Terminal-Bench 2.1 (Terminal-Bench 2.1)

> State-of-the-art set of difficult terminal-based tasks

Canonical page: https://benchlm.ai/benchmarks/valsterminalbench21

- Category: [external](/external)
- Last updated: September 10, 2026

Latest rankings: [the latest benchmark rankings](/benchmarks/terminal-bench-4)

## About Terminal-Bench 2.1

- Year: 2026
- Tasks: Terminal-based task execution
- Format: Accuracy score
- Difficulty: Frontier terminal-agent execution
- Paper: [Terminal-Bench 2.1](https://www.vals.ai/benchmarks/terminal-bench-2-1)

BenchLM mirrors the public Vals AI Terminal-Bench 2.1 leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

Terminal-Bench 2.1 is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (64 models)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [GPT-6 Astra](/models/gpt-6-astra) | max reasoning | OpenAI | 87.27% |
| 2 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | max reasoning | OpenAI | 85.77% |
| 3 | [Claude Fable 5.1](/models/claude-fable-5-1) | — | Anthropic | 85.02% |
| 4 | [Claude Opus 5](/models/claude-opus-5) | — | Anthropic | 84.64% |
| 5 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | high reasoning | Google | 81.27% |
| 6 | [Kimi K3](/models/kimi-k3) | — | Moonshot AI | 80.90% |
| 7 | [Claude Fable 5](/models/claude-fable) | — | Anthropic | 80.52% |
| 8 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | max reasoning | OpenAI | 79.03% |
| 9 | [Muse Spark 1.3 Max](https://www.vals.ai/models/meta_muse_spark_1_3_max) | max reasoning | Meta | 79.03% |
| 10 | [Grok 4.6](/models/grok-4-6) | high reasoning | xAI | 78.28% |
| 11 | [Gemini 3.7 Flash](/models/gemini-3-7-flash) | high reasoning | Google | 77.53% |
| 12 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | max reasoning | OpenAI | 77.53% |
| 13 | [GPT-5.5](/models/gpt-5-5) | high reasoning | OpenAI | 76.40% |
| 14 | [DeepSeek V4.1 Flash](/models/deepseek-v4-1-flash) | high reasoning | DeepSeek | 74.53% |
| 15 | [Claude Sonnet 5](/models/claude-sonnet-5) | — | Anthropic | 74.53% |
| 16 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | high reasoning | Google | 74.16% |
| 17 | [Gemini 3.6 Flash](/models/gemini-3-6-flash) | high reasoning | Google | 73.78% |
| 18 | [Muse Spark 1.3](/models/muse-spark-1-3) | xhigh reasoning | Meta | 72.28% |
| 19 | [Claude Opus 4.8](/models/claude-opus-4-8) | — | Anthropic | 71.91% |
| 20 | [GLM-5.3](/models/glm-5-3) | max reasoning | Z.AI | 71.54% |
| 21 | [Gemini 3.1 Pro Preview](https://www.vals.ai/models/google_gemini-3.1-pro-preview) | high reasoning | Google | 70.79% |
| 22 | [Claude Opus 4.8 Claude Code](https://www.vals.ai/models/anthropic_claude-opus-4-8-claude-code) | Claude Code · Claude Code | Anthropic | 69.66% |
| 23 | [Muse Spark 1.2](/models/muse-spark-1-2) | xhigh reasoning | Meta | 69.66% |
| 24 | [Muse Spark 1.1](/models/muse-spark-1-1) | xhigh reasoning | Meta | 69.29% |
| 25 | [Claude Opus 4.7](/models/claude-opus-4-7) | — | Anthropic | 68.54% |
| 26 | [Grok 4.5](/models/grok-4-5) | high reasoning | xAI | 67.79% |
| 27 | [GLM-5.2](/models/glm-5-2) | max reasoning | Z.AI | 67.79% |
| 28 | [Qwen3.8 Max](/models/qwen3-8-max) | — | Alibaba | 67.42% |
| 29 | [DeepSeek V4 Flash 0731](/models/deepseek-v4-flash-0731) | — | DeepSeek | 67.04% |
| 30 | [Kimi K2.7 Code](/models/kimi-k2-7-code) | — | Moonshot AI | 67.04% |
| 31 | [GLM-5.3-Flash](/models/glm-5-3-flash) | max reasoning | Z.AI | 62.92% |
| 32 | [Qwen3.7 Max](/models/qwen3-7-max) | — | Alibaba | 61.05% |
| 33 | [MiMo-V2.5](/models/mimo-v2-5) | — | Xiaomi | 60.67% |
| 34 | [Composer 2.5](/models/composer-2-5) | Cursor CLI · Cursor CLI | Cursor | 58.43% |
| 35 | [Qwen3.8-27B](/models/qwen3-8-27b) | xhigh reasoning | Alibaba | 58.43% |
| 36 | [GPT-5.5 Codex](https://www.vals.ai/models/openai_gpt-5.5-codex) | Codex · Codex | OpenAI | 57.30% |
| 37 | [GPT-5.5 Factory](https://www.vals.ai/models/openai_gpt-5.5-factory) | Factory · Factory | OpenAI | 57.30% |
| 38 | [Claude Sonnet 4.6](/models/claude-sonnet-4-6) | — | Anthropic | 57.30% |
| 39 | [MiMo-V2.5-Pro](/models/mimo-v2-5-pro) | — | Xiaomi | 57.30% |
| 40 | [GLM-5.1](/models/glm-5-1) | — | Z.AI | 56.93% |
| 41 | [Inkling-Small](/models/inkling-small) | 0.99 reasoning | Thinking Machines Lab | 55.06% |
| 42 | [GPT-5.4 mini](/models/gpt-5-4-mini) | high reasoning | OpenAI | 54.68% |
| 43 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | max reasoning | DeepSeek | 54.68% |
| 44 | [Gemini 3 Flash Preview](https://www.vals.ai/models/google_gemini-3-flash-preview) | high reasoning | Google | 53.93% |
| 45 | [Kimi K2.6](/models/kimi-2-6) | — | Moonshot AI | 53.56% |
| 46 | [MiniMax M3](/models/minimax-m3) | — | MiniMax | 53.56% |
| 47 | [Qwen3.6 Plus](/models/qwen3-6-plus) | — | Alibaba | 53.18% |
| 48 | [Qwen3.7 Plus](/models/qwen3-7-plus) | — | Alibaba | 52.81% |
| 49 | [Nemotron 3 Ultra 550b A55b](https://www.vals.ai/models/nvidia_nemotron-3-ultra-550b-a55b) | — | Nvidia | 50.94% |
| 50 | [Gemini 3.5 Flash-Lite](/models/gemini-3-5-flash-lite) | high reasoning | Google | 50.19% |
| 51 | [Ling 3.0 Flash 2607](https://www.vals.ai/models/ant_ling-3.0-flash-2607) | — | Ant | 50.19% |
| 52 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | — | DeepSeek | 50.19% |
| 53 | [MiniMax M2.7](/models/minimax-m2-7) | — | MiniMax | 48.69% |
| 54 | [Inkling](/models/inkling) | 0.99 reasoning | Thinking Machines Lab | 47.57% |
| 55 | [Grok 4.20 0309 Reasoning](https://www.vals.ai/models/grok_grok-4.20-0309-reasoning) | — | xAI | 44.20% |
| 56 | [Claude Haiku 4.5 Thinking](/models/claude-haiku-4-5-thinking) | — | Anthropic | 43.82% |
| 57 | [Grok 4.3](/models/grok-4-3) | high reasoning | xAI | 41.95% |
| 58 | [Kimi K2.5 Thinking](https://www.vals.ai/models/kimi_kimi-k2.5-thinking) | — | Moonshot AI | 41.95% |
| 59 | [GPT-5.4 nano](/models/gpt-5-4-nano) | high reasoning | OpenAI | 41.57% |
| 60 | [Mistral Medium 3.5](https://www.vals.ai/models/mistralai_mistral-medium-3.5) | high reasoning | Mistral AI | 38.95% |
| 61 | [Gemini 3.1 Flash Lite Preview](https://www.vals.ai/models/google_gemini-3.1-flash-lite-preview) | high reasoning | Google | 34.08% |
| 62 | [Laguna M.1](/models/laguna-m-1) | — | Poolside | 34.08% |
| 63 | [Laguna XS.2](/models/laguna-xs-2) | — | Poolside | 25.84% |
| 64 | [Nemotron Lightning 3p5 30b A3b](https://www.vals.ai/models/fireworks_nemotron-lightning-3p5-30b-a3b) | — | Fireworks AI | 10.86% |

## FAQ

### What does Terminal-Bench 2.1 measure?

State-of-the-art set of difficult terminal-based tasks

### Which model leads the published Terminal-Bench 2.1 snapshot?

GPT-6 Astra currently leads the published Terminal-Bench 2.1 snapshot with a score of 87.27%.

### How many models are evaluated on Terminal-Bench 2.1?

The September 10, 2026 contains 64 AI models.

### Does Terminal-Bench 2.1 affect BenchLM's overall score?

Not directly. Terminal-Bench 2.1 is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.