Terminal-Bench

Source-qualified evaluations, report statistics and safety observations.

anthropic/claude-fable-5

MetricValueUnit
Terminal-Bench · accuracy0.44549999999999995percent
Terminal-Bench · avg_trial_duration_sec4202.8seconds
Terminal-Bench · pass_at_20.5727percent
Terminal-Bench · pass_at_30.6318percent
Terminal-Bench · pass_at_40.6636percent
Terminal-Bench · pass_at_50.6818percent
Terminal-Bench · total_cost_usd7265.01USD

anthropic/claude-opus-4-8

MetricValueUnit
Terminal-Bench · accuracy0.2364percent
Terminal-Bench · avg_trial_duration_sec5163seconds
Terminal-Bench · pass_at_20.3455percent
Terminal-Bench · pass_at_30.4045percent
Terminal-Bench · pass_at_40.4424percent
Terminal-Bench · pass_at_50.4697percent
Terminal-Bench · total_cost_usd6481.26USD

anthropic/claude-opus-5

MetricValueUnit
Terminal-Bench · accuracy0.5182percent
Terminal-Bench · avg_trial_duration_sec4792.8seconds
Terminal-Bench · pass_at_20.6167percent
Terminal-Bench · pass_at_30.6576percent
Terminal-Bench · pass_at_40.6818percent
Terminal-Bench · pass_at_50.697percent
Terminal-Bench · total_cost_usd5969.11USD

anthropic/claude-sonnet-5

MetricValueUnit
Terminal-Bench · accuracy0.1242percent
Terminal-Bench · avg_trial_duration_sec6510.3seconds
Terminal-Bench · pass_at_20.2045percent
Terminal-Bench · pass_at_30.2621percent
Terminal-Bench · pass_at_40.3091percent
Terminal-Bench · pass_at_50.3485percent
Terminal-Bench · total_cost_usd9603.86USD

anthropic/glm-5.3

MetricValueUnit
Terminal-Bench · accuracy0.4182percent
Terminal-Bench · avg_trial_duration_sec5831seconds
Terminal-Bench · pass_at_20.5076percent
Terminal-Bench · pass_at_30.547percent
Terminal-Bench · pass_at_40.5667percent
Terminal-Bench · pass_at_50.5758percent
Terminal-Bench · total_cost_usd2727.63USD

openai/gpt-5.6-luna

MetricValueUnit
Terminal-Bench · accuracy0.1727percent
Terminal-Bench · avg_trial_duration_sec4088.3seconds
Terminal-Bench · pass_at_20.2424percent
Terminal-Bench · pass_at_30.2848percent
Terminal-Bench · pass_at_40.3121percent
Terminal-Bench · pass_at_50.3333percent
Terminal-Bench · total_cost_usd346.67USD

openai/gpt-5.6-sol

MetricValueUnit
Terminal-Bench · accuracy0.37270000000000003percent
Terminal-Bench · avg_trial_duration_sec2388.4seconds
Terminal-Bench · pass_at_20.4955percent
Terminal-Bench · pass_at_30.55percent
Terminal-Bench · pass_at_40.5818percent
Terminal-Bench · pass_at_50.6061percent
Terminal-Bench · total_cost_usd2541.7USD

openai/gpt-5.6-terra

MetricValueUnit
Terminal-Bench · accuracy0.2152percent
Terminal-Bench · avg_trial_duration_sec2514.6seconds
Terminal-Bench · pass_at_20.3061percent
Terminal-Bench · pass_at_30.3636percent
Terminal-Bench · pass_at_40.4061percent
Terminal-Bench · pass_at_50.4394percent
Terminal-Bench · total_cost_usd1733.52USD

xai/grok-4.5

MetricValueUnit
Terminal-Bench · accuracy0.1242percent
Terminal-Bench · avg_trial_duration_sec3162.1seconds
Terminal-Bench · pass_at_20.1833percent
Terminal-Bench · pass_at_30.2227percent
Terminal-Bench · pass_at_40.2515percent
Terminal-Bench · pass_at_50.2727percent
Terminal-Bench · total_cost_usd2094.11USD

xai/grok-4.6

MetricValueUnit
Terminal-Bench · accuracy0.203percent
Terminal-Bench · avg_trial_duration_sec2351.5seconds
Terminal-Bench · pass_at_20.2848percent
Terminal-Bench · pass_at_30.3303percent
Terminal-Bench · pass_at_40.3636percent
Terminal-Bench · pass_at_50.3939percent
Terminal-Bench · total_cost_usd3591.58USD

gemini/gemini-3.8-flash

MetricValueUnit
Terminal-Bench · accuracy0.1909percent
Terminal-Bench · avg_trial_duration_sec1995.1seconds
Terminal-Bench · pass_at_20.2879percent
Terminal-Bench · pass_at_30.353percent
Terminal-Bench · pass_at_40.4percent
Terminal-Bench · pass_at_50.4394percent
Terminal-Bench · total_cost_usd0USD

anthropic/claude-fable-5-1

MetricValueUnit
Terminal-Bench · accuracy0.5788percent
Terminal-Bench · avg_trial_duration_sec3893.5seconds
Terminal-Bench · pass_at_20.7percent
Terminal-Bench · pass_at_30.747percent
Terminal-Bench · pass_at_40.7727percent
Terminal-Bench · pass_at_50.7879percent
Terminal-Bench · total_cost_usd6243.5USD

gemini/gemini-3.7-flash

MetricValueUnit
Terminal-Bench · accuracy0.1121percent
Terminal-Bench · avg_trial_duration_sec1688.9seconds
Terminal-Bench · pass_at_20.1636percent
Terminal-Bench · pass_at_30.2percent
Terminal-Bench · pass_at_40.2303percent
Terminal-Bench · pass_at_50.2576percent
Terminal-Bench · total_cost_usd1261.87USD

anthropic/claude-opus-5-5

MetricValueUnit
Terminal-Bench · accuracy0.6485percent
Terminal-Bench · avg_trial_duration_sec4790.4seconds
Terminal-Bench · pass_at_20.7318percent
Terminal-Bench · pass_at_30.7652percent
Terminal-Bench · pass_at_40.7788percent
Terminal-Bench · pass_at_50.7879percent
Terminal-Bench · total_cost_usd4718.65USD

anthropic/claude-sonnet-5-5

MetricValueUnit
Terminal-Bench · accuracy0.6182percent
Terminal-Bench · avg_trial_duration_sec7431.6seconds
Terminal-Bench · pass_at_20.6924percent
Terminal-Bench · pass_at_30.7364percent
Terminal-Bench · pass_at_40.7606percent
Terminal-Bench · pass_at_50.7727percent
Terminal-Bench · total_cost_usd7333.64USD

meta/muse-spark-1.3

MetricValueUnit
Terminal-Bench · accuracy0.14550000000000002percent
Terminal-Bench · avg_trial_duration_sec4341.7seconds
Terminal-Bench · pass_at_20.2197percent
Terminal-Bench · pass_at_30.2652percent
Terminal-Bench · pass_at_40.3percent
Terminal-Bench · pass_at_50.3333percent
Terminal-Bench · total_cost_usd0USD

openai/gpt-6.1-sol

MetricValueUnit
Terminal-Bench · accuracy0.5818percent
Terminal-Bench · avg_trial_duration_sec3176.7seconds
Terminal-Bench · pass_at_20.6652percent
Terminal-Bench · pass_at_30.7015percent
Terminal-Bench · pass_at_40.7242percent
Terminal-Bench · pass_at_50.7424percent
Terminal-Bench · total_cost_usd634.24USD

openai/gpt-6-luna

MetricValueUnit
Terminal-Bench · accuracy0.1636percent
Terminal-Bench · avg_trial_duration_sec2284.1seconds
Terminal-Bench · pass_at_20.2273percent
Terminal-Bench · pass_at_30.2621percent
Terminal-Bench · pass_at_40.2848percent
Terminal-Bench · pass_at_50.303percent
Terminal-Bench · total_cost_usd72.14USD

Models

Loading
EVIDENCE / COMPARISON

Benchmarks

Explore every benchmark

Loading

These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

Install TensorTrace