EleutherAI: Llemma 7b

Text input Text output
Author's Description

Llemma 7B is a language model for mathematics. It was initialized with Code Llama 7B weights, and trained on the Proof-Pile-2 for 200B tokens. Llemma models are particularly strong at chain-of-thought mathematical reasoning and using computational tools for mathematics, such as Python and formal theorem provers.

Key Specifications
Cost
$$$$
Context
4K
Parameters
7B
Released
Apr 14, 2025
Speed
Ability
Reliability
Supported Parameters

This model supports the following parameters:

Stop Presence Penalty Top P Temperature Min P Seed Frequency Penalty Max Tokens
Performance Summary

EleutherAI's Llemma 7B, a language model specialized in mathematics, demonstrates exceptional performance in terms of speed and cost-efficiency. It consistently ranks among the fastest models and offers highly competitive pricing across all evaluated benchmarks, achieving an "Infinityth percentile" in both categories. This indicates a significant advantage in operational efficiency. However, despite its strong foundation in mathematical reasoning and tool use, Llemma 7B exhibits notable weaknesses in general-purpose tasks. Its accuracy across benchmarks like Coding (4.0%), Instruction Following (0.0%), Email Classification (8.0%), Ethics (7.0%), and General Knowledge (7.5%) is very low, often falling into the lowest percentiles. The model's primary strength appears to be in Reasoning, where it achieved 50.0% accuracy, placing it in the 38th percentile, which is still moderate. The description highlights its strength in mathematical reasoning and tool use, which aligns with the relatively better performance in the general "Reasoning" category, though specific mathematical benchmarks are not provided. Its reliability, as indicated by the duration percentiles, is generally very low across most benchmarks, suggesting potential issues with consistently providing timely responses.

Model Pricing

Current Pricing

Feature Price (per 1M tokens)
Prompt $0.8
Completion $1.2

Price History

Available Endpoints
Provider Endpoint Name Context Length Pricing (Input) Pricing (Output)
Featherless
Featherless | eleutherai/llemma_7b 4K $0.8 / 1M tokens $1.2 / 1M tokens
Benchmark Results
Benchmark Category Reasoning Free Executions Accuracy Cost Duration