Command Light: Complete Benchmark Performance Guide

cohere light 2023-09-14 CC-BY-NC-4.0

In-depth analysis of Command Light performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorcohere
Versionlight
Release Date2023-09-14
Context Window4000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseCC-BY-NC-4.0
Documentationhttps://docs.cohere.com/docs

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 52.8 % 2023-09-14 5-shot view
HUMANEVAL 51.2 pass@1 2023-09-14 — view
GSM8K 39.8 % 2023-09-14 0-shot CoT view
MATH 24.1 % 2023-09-14 0-shot CoT view
BBH 55.5 % 2023-09-14 3-shot CoT view
GPQA 30.1 % 2023-09-14 0-shot view
IFEVAL 53.9 % 2023-09-14 prompt_strict view
ARC 85.9 % 2023-09-14 challenge view
MUSR 40.7 % 2023-09-14 0-shot view
WINOGRANDE 74.7 % 2023-09-14 0-shot view

Pricing

Tier Price Currency
Input$0.5 / MtokUSD
Output$0.5 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://cohere.com/pricing · as of 2023-09-14

Compliance

  • Data Residency: US
  • SOC2: ✓
  • HIPAA: ✗
  • GDPR: ✓
  • ISO 27001: ✓

Command Light

Model Overview

Cohere Command Light 经济型模型, 4K 上下文, 速度快成本低, 适合简单任务与高吞吐场景。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Cohere light 2023-09-14 4K text text CC-BY-NC-4.0

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 52.8 % 5-shot
HumanEval 51.2 pass@1 —
GSM8K (Grade School Math 8K) 39.8 % 0-shot CoT
MATH 24.1 % 0-shot CoT
BBH (BIG-Bench Hard) 55.5 % 3-shot CoT
GPQA 30.1 % 0-shot
IFEval 53.9 % prompt_strict
ARC 85.9 % challenge
MUSR 40.7 % 0-shot
WinoGrande 74.7 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • MMLU 52.8, weak knowledge reasoning.
  • Proprietary, not self-hostable.
  • Context window 4K is limited.

Use Cases

  • General chat and Q&A

References