bench view

View and analyze previous evaluation results

Overview

The bench view command displays results from previous evaluations, allowing you to analyze performance, compare models, and track progress over time.

Usage

bench view [options]

Options

Option	Description
--log-dir	Log directory to view (defaults to ./logs)
--recursive —no-recursive`	Include all logs in log_dir recursively
--host	CP/IP host for server
--port	TCP/IP port for server
--log-level	Set the log level

Examples

View Latest Result

bench view

Example Evaluation Logs Summary:

localhost

| Task      |       Model        | Score | Status |         Completed        |                            File Name                           |
|-----------|--------------------|-------|--------|--------------------------|----------------------------------------------------------------|
| mmlu      | openai/o3-mini     | 0.82  |    ✓   | Sat Aug 16 2025 10:39 PM | 2025-08-16T22-39-13-04-00_mmlu_g5QsKYFFAR7zNSuMMs9a85.eval     |
| humaneval | anthropic/claude-3 | 0.74  |    ✓   | Fri Aug 16 2025 03:22 PM | 2025-08-16T15-22-41-08-00_humaneval_k2mNpR8vLx3wQfE7Hs4B2.eval |
| gpqa      | groq/llama-3.3-70b | 0.43  |    ✓   | Thu Aug 04 2025 11:45 AM | 2025-08-04T11-45-09-12-00_gpqa_diamond_v9XzTpL5Kj8rY3mQ7.eval |
| math      | openai/gpt-4o      | 0.67  |    ✓   | Wed Aug 03 2025 08:15 AM | 2025-08-03T08-15-32-07-00_math_u4JhWq2NvL6xKc9PzM8sA1.eval    |
| simpleqa  | openai/gpt-4o-mini | 0.58  |    ⚠   | Tue Jul 07 2025 05:30 PM | 2025-07-07T17-30-18-05-00_simpleqa_b7FgRp3XvK2nY9jQ6L.eval    |
| ...       | ...                | ...   |  ...   | ...                      | ...                                                            |

Each entry in the evaluation logs summary can be expanded to show a detailed evaluation breakdown: