Terminal Bench 4
An AI model benchmark focused on complex task automation in command-line environments. It includes scenarios involving system tool invocation, file handling, and multi-step operations, placing high demands on a model's practical engineering deployment capabilities.
Core Facts
On Terminal-Bench 4.0, Claude Opus 5.5 leads with 66.4%, followed by GPT-6 Astra at 61.8% and Argon at 57.4%.
Claude Sonnet 5.5 achieved 70.6% on the Terminal benchmark, compared to only 10.3% for the previous generation.
On the Terminal Bench terminal task test, Gemini 4 Argon scored 57.4%, below Claude Opus 5.5's 66.4%.
Timeline (last 90 days)
Running Terminal Bench 4 under Val's unified AI agent, Argon scored 57.6, Opus 65.2, and Sonnet 64.1.
On the Terminal Bench terminal task test, Gemini 4 Argon scored 57.4%, below Claude Opus 5.5's 66.4%.
On Terminal-Bench 4.0, Claude Opus 5.5 leads with 66.4%, followed by GPT-6 Astra at 61.8% and Argon at 57.4%.
Claude Sonnet 5.5 achieved 70.6% on the Terminal benchmark, compared to only 10.3% for the previous generation.
On Terminal Bench 4, Opus 5.5's medium tier scored higher than Fable's max tier, at less than half the price (Fable max tier at $20, Opus 5.5 medium tier at $2.94).
All Facts (5)
On Terminal-Bench 4.0, Claude Opus 5.5 leads with 66.4%, followed by GPT-6 Astra at 61.8% and Argon at 57.4%.
75%VerifiedClaude Sonnet 5.5 achieved 70.6% on the Terminal benchmark, compared to only 10.3% for the previous generation.
75%Partially VerifiedOn the Terminal Bench terminal task test, Gemini 4 Argon scored 57.4%, below Claude Opus 5.5's 66.4%.
65%UnverifiedRunning Terminal Bench 4 under Val's unified AI agent, Argon scored 57.6, Opus 65.2, and Sonnet 64.1.
50%UnverifiedOn Terminal Bench 4, Opus 5.5's medium tier scored higher than Fable's max tier, at less than half the price (Fable max tier at $20, Opus 5.5 medium tier at $2.94).
50%