Llemma 7B: Complete Benchmark Performance Guide

other llemma-7b 2023-10-17 Apache 2.0

In-depth analysis of Llemma 7B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionllemma-7b
Release Date2023-10-17
Context Window4096 tokens
Input Modalitiestext
Output Modalitiestext
LicenseApache 2.0
Documentationhttps://huggingface.co/models

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 66 % 2023-10-17 5-shot view
HUMANEVAL 64.2 pass@1 2023-10-17 — view
GSM8K 84.9 % 2023-10-17 0-shot CoT view
MATH 47.5 % 2023-10-17 0-shot CoT view
BBH 73.1 % 2023-10-17 3-shot CoT view
GPQA 37.4 % 2023-10-17 0-shot view
IFEVAL 52.9 % 2023-10-17 prompt_strict view
ARC 87.4 % 2023-10-17 challenge view
MUSR 42.6 % 2023-10-17 0-shot view
WINOGRANDE 70.8 % 2023-10-17 0-shot view

Pricing

Tier Price Currency
Input$0.18 / MtokUSD
Output$0.18 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2023-10-17

Compliance

  • Data Residency: self-host
  • SOC2: ✗
  • HIPAA: ✗
  • GDPR: ✗
  • ISO 27001: ✗

Llemma 7B

Model Overview

TheBloke Llemma 7B 数学专用模型, 4K 上下文, 基于 Code Llama 7B 继续训练, 数学推理能力突出。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Other llemma-7b 2023-10-17 4K text text Apache 2.0

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 66.0 % 5-shot
HumanEval 64.2 pass@1 —
GSM8K (Grade School Math 8K) 84.9 % 0-shot CoT
MATH 47.5 % 0-shot CoT
BBH (BIG-Bench Hard) 73.1 % 3-shot CoT
GPQA 37.4 % 0-shot
IFEval 52.9 % prompt_strict
ARC 87.4 % challenge
MUSR 42.6 % 0-shot
WinoGrande 70.8 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • Proprietary, not self-hostable.
  • Context window 4K is limited.

Use Cases

  • General chat and Q&A

References