Capybara 1.5B: Complete Benchmark Performance Guide

other capybara-1-5b 2023-11-15 Apache 2.0

In-depth analysis of Capybara 1.5B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versioncapybara-1-5b
Release Date2023-11-15
Context Window4096 tokens
Input Modalitiestext
Output Modalitiestext
LicenseApache 2.0
Documentationhttps://huggingface.co/models

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 51.7 % 2023-11-15 5-shot view
HUMANEVAL 30.1 pass@1 2023-11-15 — view
GSM8K 28.5 % 2023-11-15 0-shot CoT view
MATH 25.6 % 2023-11-15 0-shot CoT view
BBH 48.3 % 2023-11-15 3-shot CoT view
GPQA 18.1 % 2023-11-15 0-shot view
IFEVAL 52.4 % 2023-11-15 prompt_strict view
ARC 80.1 % 2023-11-15 challenge view
MUSR 34 % 2023-11-15 0-shot view
WINOGRANDE 68.9 % 2023-11-15 0-shot view

Pricing

Tier Price Currency
Input$0.08 / MtokUSD
Output$0.08 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2023-11-15

Compliance

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

Capybara 1.5B

Model Overview

IntrinsicaAI Capybara 1.5B 轻量对话模型, 4K 上下文, 1.5B 参数, 适合边缘设备对话场景。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Other capybara-1-5b 2023-11-15 4K text text Apache 2.0

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 51.7 % 5-shot
HumanEval 30.1 pass@1 —
GSM8K (Grade School Math 8K) 28.5 % 0-shot CoT
MATH 25.6 % 0-shot CoT
BBH (BIG-Bench Hard) 48.3 % 3-shot CoT
GPQA 18.1 % 0-shot
IFEval 52.4 % prompt_strict
ARC 80.1 % challenge
MUSR 34.0 % 0-shot
WinoGrande 68.9 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • MMLU 51.7, weak knowledge reasoning.
  • HumanEval 30.1, coding weak.
  • Proprietary, not self-hostable.
  • Context window 4K is limited.

Use Cases

  • General chat and Q&A

References