LLM ReferenceLLM Reference

OpenHermes 2.5 Mistral 7B

About

OpenHermes 2.5 Mistral 7B is an advanced large language model developed by Teknium, building on the previous version, OpenHermes 2. Utilizing a transformer architecture, it's fine-tuned on over one million entries, combining code and non-code data, primarily composed of GPT-4 generated text. This enhances its human-like response capabilities across diverse contexts. It excels in conversational AI with its multi-turn dialogue support through the ChatML format, significantly improves in code generation tasks with a high HumanEval score, and performs robustly on benchmarks like GPT4All and AGIEval. Additionally, it offers various quantization options for optimized hardware performance and is adept at real-time data processing, making it ideal for customer service and immediate data analysis applications.

Capabilities

VisionMultimodalReasoningFunction CallingTool UseStructured OutputsCode Execution

Providers(3)

Compare all →
ProviderInput (per 1M)Output (per 1M)Type
Together AI$0.2$0.2Serverless
Cloudflare Workers AIServerless
Fireworks AI$0.20$0.20Provisioned

Benchmark Scores(4)

BenchmarkScoreVersionSource
Google-Proof Q&A38.6diamondOpen LLM Leaderboard
HellaSwag86.010-shotOpen LLM Leaderboard
HumanEval54.8pass@1Open LLM Leaderboard
Massive Multitask Language Understanding63.55-shotOpen LLM Leaderboard

Rankings

Specifications

Released2023-12-15
Parameters7B
Context32K
ArchitectureDecoder Only
Knowledge cutoff2023-12
Specializationgeneral
Trainingfinetuning

Created by

Steerable AI models for open-source innovation

N/A
Founded N/A
Website