Which AI model scores highest on Terminal-Bench 3.0?

DeepSeek-V4.1-Flash by DeepSeek holds the best Terminal-Bench 3.0 result among tracked models, at 30% (released Sep 10 2026). Higher scores are better on this benchmark.

What are the top 5 models on Terminal-Bench 3.0?

1. DeepSeek-V4.1-Flash (DeepSeek) — 30%; 2. Qwen3.8-Max-0902 (Qwen) — 29%; 3. GLM-5.3 (Z.ai) — 28.3%; 4. Grok 4.6 (SpaceXAI) — 26%; 5. Grok 4.5 (SpaceXAI) — 15.7%.

How many models have a published Terminal-Bench 3.0 score?

7 tracked models have a published Terminal-Bench 3.0 score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.

What is the best open model on Terminal-Bench 3.0?

GLM-5.3 by Z.ai is the highest-ranked model with downloadable weights on Terminal-Bench 3.0, scoring 28.3% at rank 3 overall.

Agentic terminal coding

Terminal-Bench 3.0

command-line task completion (v3.0, much harder task set)

Rankings

Higher is better

Terminal-Bench 3.0 — frequently asked questions

: command-line task completion (v3.0, much harder task set)
: DeepSeek-V4.1-Flash by DeepSeek holds the best Terminal-Bench 3.0 result among tracked models, at 30% (released Sep 10 2026). Higher scores are better on this benchmark.
: 1. DeepSeek-V4.1-Flash (DeepSeek) — 30%; 2. Qwen3.8-Max-0902 (Qwen) — 29%; 3. GLM-5.3 (Z.ai) — 28.3%; 4. Grok 4.6 (SpaceXAI) — 26%; 5. Grok 4.5 (SpaceXAI) — 15.7%.
: 7 tracked models have a published Terminal-Bench 3.0 score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.
: GLM-5.3 by Z.ai is the highest-ranked model with downloadable weights on Terminal-Bench 3.0, scoring 28.3% at rank 3 overall.